Sorbonne Université, Deep Leaning Project, Mars 2020 : Charly Nana Konguep

COMPUTER VISION PROJECT

Image classification with Tensorflow and Keras

Introduction

In this project, we are interested in the classification of images into two classes. The images of the study are images of the real world, complex to classify, because all the images do not have the same size, the same shape, they are not monochrome or even centered. In addition, on some images, we observe humans and various backgrounds: which complicates the task of classification for the models.

The database consists of 3000 image files divided into two folders. A cats folder made up of 1500 cat images and adogs folder made up of 1500 dog images.

Project Structure

Task #1: Import Libraries and Data Preprocessing

Task #2: Visualize Examples

Task #3: Create model from scratch

Task #4: Data Preprocessing : Build Image Data Generator

Task #5: Train Model

Task #6: Visualization of the Accuracy and Loss

Task #7: Data Augmentation and Re-training of the model

Task #8: Transfert Learning with Data Augmentation

Task #9: Assess the performance of the trained model

Task #1 : Import Libraries and Data Preprocessing

• Import cats_vs_dogs dataset
• Split the dataset into training and validation data.
• Randomly shuffling the newly created subset

Import cats_vs_dogs dataset

import os
import random
import shutil
from shutil import copyfile

CAT_SOURCE_DIR = os.path.join('./cats/')
DOG_SOURCE_DIR = os.path.join('./dogs/')

print('Total cats images :',len(os.listdir(CAT_SOURCE_DIR)))
print('Total dogs images :', len(os.listdir(DOG_SOURCE_DIR)))

Total cats images : 1500
Total dogs images : 1500

Data Splitting

# We use os.mkdir to create our directories.
# We need two directories for training and testing, and subdirectories for cats and dogs.
try:
    os.mkdir('./training/')
    os.mkdir('./training/cats/')
    os.mkdir('./training/dogs/')
    
    os.mkdir('./testing/')
    os.mkdir('./testing/cats/')
    os.mkdir('./testing/dogs/')

except OSError:
    pass

# Directories for training data
TRAINING_DIR = os.path.join('./training/')

# Directories for testing data
TESTING_DIR = os.path.join('./testing/')

# Subdirectories for training data
TRAINING_CATS_DIR = os.path.join('./training/cats/')
TRAINING_DOGS_DIR = os.path.join('./training/dogs/')

# Subdirectories for testing data
TESTING_CATS_DIR = os.path.join('./testing/cats/')
TESTING_DOGS_DIR = os.path.join('./testing/dogs/')

def split_data(SOURCE, TRAINING, TESTING, SPLIT_SIZE = 0.8):
    """
    Splitting data for training and testing data.
    
    - SOURCE : directory containing the files (cats or dogs)
    - TRAINING : Directory in which a portion of the SOURCE file will be copied. 
    - TESTING : Directory where the rest of the SOURCE file will be copied. 
    - SPLIT SIZE : real between 0 and 1, to determine the portion that will be copied in TRAINING and the rest in TESTING
    
    """
    file_names = os.listdir(SOURCE)
    file_names_shuffle = random.sample(file_names, len(file_names))
    n = int(len(file_names) * SPLIT_SIZE)
    training_list = file_names_shuffle[:n]
    testing_list = file_names_shuffle[n:]
    
    for file1 in training_list:
        shutil.copyfile(os.path.join(SOURCE, file1), os.path.join(TRAINING, file1))
        
    for file2 in testing_list:
        shutil.copyfile(os.path.join(SOURCE, file2), os.path.join(TESTING, file2))
        
# We split data    
split_data(CAT_SOURCE_DIR, TRAINING_CATS_DIR, TESTING_CATS_DIR, SPLIT_SIZE = 0.9)
split_data(DOG_SOURCE_DIR, TRAINING_DOGS_DIR, TESTING_DOGS_DIR, SPLIT_SIZE = 0.9)

print('Total training cats :', len(os.listdir(TRAINING_CATS_DIR)))
print('Total training dogs :', len(os.listdir(TRAINING_DOGS_DIR)))

print('Total testing cats :', len(os.listdir(TESTING_CATS_DIR)))
print('Total testing dogs :', len(os.listdir(TESTING_DOGS_DIR)))

Total training cats : 1350
Total training dogs : 1350
Total testing cats : 150
Total testing dogs : 150

os.listdir(TRAINING_DIR)

['cats', 'dogs']

os.listdir(os.path.join(TRAINING_DIR, 'dogs'))[ : 10]

['dog.0.jpg',
 'dog.1.jpg',
 'dog.100.jpg',
 'dog.101.jpg',
 'dog.102.jpg',
 'dog.103.jpg',
 'dog.104.jpg',
 'dog.105.jpg',
 'dog.107.jpg',
 'dog.11.jpg']

os.listdir(os.path.join(TRAINING_DIR, 'cats'))[ : 10]

['cat.0.jpg',
 'cat.1.jpg',
 'cat.10.jpg',
 'cat.100.jpg',
 'cat.101.jpg',
 'cat.102.jpg',
 'cat.103.jpg',
 'cat.104.jpg',
 'cat.105.jpg',
 'cat.106.jpg']

# Check the number of images in the dataset
train_file_path = []
label_names = []

# os.listdir returns the list of files in the folder, in this case image class names
for i in  os.listdir(TRAINING_DIR):
  train_class = os.listdir(os.path.join(TRAINING_DIR, i))
  for j in train_class:
    img = os.path.join(TRAINING_DIR, i, j)
    train_file_path.append(img)
    label_names.append(i)

print('Number of train images : {} \n'.format(len(train_file_path)))

Number of train images : 2700 

train_file_path[ :10]

['./training/cats\\cat.0.jpg',
 './training/cats\\cat.1.jpg',
 './training/cats\\cat.10.jpg',
 './training/cats\\cat.100.jpg',
 './training/cats\\cat.101.jpg',
 './training/cats\\cat.102.jpg',
 './training/cats\\cat.103.jpg',
 './training/cats\\cat.104.jpg',
 './training/cats\\cat.105.jpg',
 './training/cats\\cat.106.jpg']

label_names[ :10]

['cats',
 'cats',
 'cats',
 'cats',
 'cats',
 'cats',
 'cats',
 'cats',
 'cats',
 'cats']

# Create Dataframe
import pandas as pd
df = pd.DataFrame({'Images': train_file_path,'Labels': label_names})
df

	Images	Labels
0	./training/cats\cat.0.jpg	cats
1	./training/cats\cat.1.jpg	cats
2	./training/cats\cat.10.jpg	cats
3	./training/cats\cat.100.jpg	cats
4	./training/cats\cat.102.jpg	cats
...	...	...
2695	./training/dogs\dog.994.jpg	dogs
2696	./training/dogs\dog.995.jpg	dogs
2697	./training/dogs\dog.996.jpg	dogs
2698	./training/dogs\dog.998.jpg	dogs
2699	./training/dogs\dog.999.jpg	dogs

2700 rows × 2 columns

import seaborn as sns

sns.countplot(x = label_names)

<AxesSubplot:ylabel='count'>

Task #2: Visualize Examples

• Plotting randomly selected examples of a given set.
• We look at some examples from training and test set along with their labels.

%matplotlib inline

import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import numpy as np

pic_index = 10
cats = np.random.permutation(os.listdir(TRAINING_CATS_DIR))
dogs = np.random.permutation(os.listdir(TRAINING_DOGS_DIR))

next_cats = [os.path.join(TRAINING_CATS_DIR, fname) for fname in cats[0:pic_index]]
next_dogs = [os.path.join(TRAINING_DOGS_DIR, fname) for fname in dogs[0:pic_index]]

plt.figure(figsize = (30, 30))
next_img = np.random.permutation(next_cats + next_dogs)
for i, img_path in enumerate(next_img):
    #print(img_path)
    plt.subplot(5, 4, i+1)
    img = mpimg.imread(img_path)
    plt.imshow(img)
    plt.xticks([])
    plt.yticks([])
    plt.xlabel(next_img[i][15:], color = 'green', fontsize=20)

plt.show()

Task #3 : Create model from scratch

• Creating a Keras Sequential model.
• Creating a function to add a convolutional block to the model.
• A look at the model summary.

import tensorflow as tf
from tensorflow.keras.layers import Input, Dense, Conv2D, MaxPooling2D, BatchNormalization
from tensorflow.keras.layers import Dropout, Flatten, Activation
from tensorflow.keras.optimizers import RMSprop
from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint

def create_model():
    def add_conv_block(model, num_filters):
        model.add(Conv2D(filters = num_filters, kernel_size = (3,3), strides = (1, 1), activation='relu', padding = 'valid'))
        #model.add(BatchNormalization())
        model.add(MaxPooling2D(pool_size = (2, 2)))
        #model.add(Dropout(rate = 0.2))
        return model
    
    model = tf.keras.models.Sequential()
    model.add(Input(shape = (150, 150, 3)))
    
    model = add_conv_block(model, 32)
    model = add_conv_block(model, 32)
    model = add_conv_block(model, 64)
    model = add_conv_block(model, 64)
    
    model.add(Flatten())
    model.add(Dense(units = 512, activation = 'relu'))
    #model.add(Dropout(rate = 0.2))
    model.add(Dense(units = 1, activation = 'sigmoid'))
    
    # Compile the model
    model.compile(optimizer=RMSprop(lr=1e-4), loss='binary_crossentropy', metrics = ['accuracy'])
    
    return model

model = create_model()
model.summary()
        

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d (Conv2D)              (None, 148, 148, 32)      896       
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 74, 74, 32)        0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 72, 72, 32)        9248      
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 36, 36, 32)        0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 34, 34, 64)        18496     
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 17, 17, 64)        0         
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 15, 15, 64)        36928     
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 7, 7, 64)          0         
_________________________________________________________________
flatten (Flatten)            (None, 3136)              0         
_________________________________________________________________
dense (Dense)                (None, 512)               1606144   
_________________________________________________________________
dense_1 (Dense)              (None, 1)                 513       
=================================================================
Total params: 1,672,225
Trainable params: 1,672,225
Non-trainable params: 0
_________________________________________________________________

Task #4: Data Preprocessing

• Buid image generator for training and testing set
• Normalizes or rescale data
• Initialize the bacth size
• Using Callback to control training

#using early stopping to exit training if validation loss is not decreasing even after certain epochs (patience)
earlystopping = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=15)

#save the best model with lower validation loss
#checkpointer = ModelCheckpoint(filepath="weights_.hdf5", verbose=1, save_best_only=True)

# Using Callback to control training
class myCallback(tf.keras.callbacks.Callback):
    def on_epoch_end(self, epoch, log={}):
        if(log.get('val_accuracy') >= 0.98):
            print("\nReached {}% val_accuracy so cancelling training!".format(round(100 * log.get('val_accuracy'), 2)))
            self.model.stop_training = True
            
# callback instance
callbacks = myCallback()

from tensorflow.keras.preprocessing.image import ImageDataGenerator

# Create run-time augmentation on training and test dataset
# For training datagenerator, we add normalization and validation_split
train_datagen = ImageDataGenerator(rescale = 1 / 255.0,
                                   validation_split = 0.1112)

# For test datagenerator, we only normalize the data.
test_datagen = ImageDataGenerator(rescale=1 / 255.0)

# Flow training images in batches of 128 using train_datagen generator
training_generator = train_datagen.flow_from_directory(
    TRAINING_DIR,                # This is the source directory for training images
    target_size = (150, 150), # All images will be resized to 150x150
    batch_size = 20,
    subset = 'training',
    class_mode = 'binary')    # Since we use binary_crossentropy loss, we need binary labels

# Flow training images in batches of 128 using train_datagen generator
validation_generator = train_datagen.flow_from_directory(
        TRAINING_DIR,  # This is the source directory for training images
        target_size = (150, 150),           # All images will be resized to 150x150
        batch_size = 20,
        subset = 'validation',
        class_mode ='binary')  # Since we use binary_crossentropy loss, we need binary labels

# Flow training images in batches of 128 using train_datagen generator
testing_generator = test_datagen.flow_from_directory(
        TESTING_DIR,  # This is the source directory for training images
        target_size = (150, 150),           # All images will be resized to 150x150
        batch_size = 20,
        class_mode ='binary')  # Since we use binary_crossentropy loss, we need binary labels

Found 2400 images belonging to 2 classes.
Found 300 images belonging to 2 classes.
Found 300 images belonging to 2 classes.

Task #5: Train Model

• Fit the model on the subset.
• Setting the EarlyStopping callback.
• Setting the ModelCheckpoint callback.

history = model.fit(
    training_generator,
    steps_per_epoch = training_generator.n // 20,
    epochs = 30,
    validation_data = validation_generator,
    validation_steps = validation_generator.n // 20,
    callbacks = [callbacks, earlystopping],
    verbose = 2
)

Epoch 1/30
120/120 - 57s - loss: 0.6848 - accuracy: 0.5508 - val_loss: 0.7097 - val_accuracy: 0.5133
Epoch 2/30
120/120 - 61s - loss: 0.6520 - accuracy: 0.6067 - val_loss: 0.6342 - val_accuracy: 0.6467
Epoch 3/30
120/120 - 60s - loss: 0.6204 - accuracy: 0.6642 - val_loss: 0.6174 - val_accuracy: 0.6500
Epoch 4/30
120/120 - 61s - loss: 0.5857 - accuracy: 0.6929 - val_loss: 0.5845 - val_accuracy: 0.6767
Epoch 5/30
120/120 - 64s - loss: 0.5593 - accuracy: 0.7092 - val_loss: 0.5583 - val_accuracy: 0.7200
Epoch 6/30
120/120 - 62s - loss: 0.5362 - accuracy: 0.7308 - val_loss: 0.5463 - val_accuracy: 0.7167
Epoch 7/30
120/120 - 66s - loss: 0.5214 - accuracy: 0.7375 - val_loss: 0.5342 - val_accuracy: 0.7467
Epoch 8/30
120/120 - 61s - loss: 0.5005 - accuracy: 0.7583 - val_loss: 0.5454 - val_accuracy: 0.7533
Epoch 9/30
120/120 - 63s - loss: 0.4813 - accuracy: 0.7696 - val_loss: 0.5121 - val_accuracy: 0.7300
Epoch 10/30
120/120 - 65s - loss: 0.4698 - accuracy: 0.7721 - val_loss: 0.5576 - val_accuracy: 0.6900
Epoch 11/30
120/120 - 66s - loss: 0.4499 - accuracy: 0.7912 - val_loss: 0.4965 - val_accuracy: 0.7767
Epoch 12/30
120/120 - 62s - loss: 0.4252 - accuracy: 0.7946 - val_loss: 0.4822 - val_accuracy: 0.7700
Epoch 13/30
120/120 - 60s - loss: 0.4109 - accuracy: 0.8121 - val_loss: 0.4736 - val_accuracy: 0.7700
Epoch 14/30
120/120 - 61s - loss: 0.3953 - accuracy: 0.8263 - val_loss: 0.4543 - val_accuracy: 0.8067
Epoch 15/30
120/120 - 66s - loss: 0.3756 - accuracy: 0.8354 - val_loss: 0.4780 - val_accuracy: 0.7867
Epoch 16/30
120/120 - 67s - loss: 0.3582 - accuracy: 0.8450 - val_loss: 0.4677 - val_accuracy: 0.7800
Epoch 17/30
120/120 - 64s - loss: 0.3441 - accuracy: 0.8454 - val_loss: 0.4839 - val_accuracy: 0.7667
Epoch 18/30
120/120 - 69s - loss: 0.3259 - accuracy: 0.8621 - val_loss: 0.4720 - val_accuracy: 0.7900
Epoch 19/30
120/120 - 72s - loss: 0.3066 - accuracy: 0.8679 - val_loss: 0.4603 - val_accuracy: 0.8000
Epoch 20/30
120/120 - 70s - loss: 0.2861 - accuracy: 0.8858 - val_loss: 0.4676 - val_accuracy: 0.7933
Epoch 21/30
120/120 - 72s - loss: 0.2737 - accuracy: 0.8913 - val_loss: 0.4659 - val_accuracy: 0.8033
Epoch 22/30
120/120 - 74s - loss: 0.2628 - accuracy: 0.8933 - val_loss: 0.4670 - val_accuracy: 0.7833
Epoch 23/30
120/120 - 66s - loss: 0.2501 - accuracy: 0.8988 - val_loss: 0.4748 - val_accuracy: 0.7967
Epoch 24/30
120/120 - 68s - loss: 0.2296 - accuracy: 0.9129 - val_loss: 0.4837 - val_accuracy: 0.7900
Epoch 25/30
120/120 - 67s - loss: 0.2133 - accuracy: 0.9233 - val_loss: 0.4826 - val_accuracy: 0.7933
Epoch 26/30
120/120 - 68s - loss: 0.1984 - accuracy: 0.9242 - val_loss: 0.6656 - val_accuracy: 0.7267
Epoch 27/30
120/120 - 70s - loss: 0.1792 - accuracy: 0.9333 - val_loss: 0.4998 - val_accuracy: 0.8000
Epoch 28/30
120/120 - 70s - loss: 0.1716 - accuracy: 0.9371 - val_loss: 0.5013 - val_accuracy: 0.7933
Epoch 29/30
120/120 - 71s - loss: 0.1556 - accuracy: 0.9413 - val_loss: 0.6092 - val_accuracy: 0.7467
Epoch 00029: early stopping

Task #6 : Visualization of the Accuracy and Loss

• Creating a function to plot training and validation accuracy per epoch, then training and validation loss
• Visualization training and validation accuracy
• Visualize training and validation loss

def graph_plot(history):
    """
    plot training and validation accuracy per epoch, then training and validation loss.
    
    Argument :
    - history : output of the trained model
    
    """
    #-----------------------------------------------------------
    # Retrieve a list of results on training and test data
    # sets for each training epoch
    #-----------------------------------------------------------
    train_acc = history.history['accuracy']
    val_acc = history.history['val_accuracy']

    train_loss = history.history['loss']
    val_loss = history.history['val_loss']

    num_epochs = range(len(train_acc)) # Get number of epochs
    
    #------------------------------------------------
    # Plot training and validation accuracy per epoch
    #------------------------------------------------
    import matplotlib.pyplot as plt
    plt.figure(figsize = (15, 4))

    plt.subplot(1, 2, 1)
    plt.plot(num_epochs, train_acc, 'bo', label='Training accuracy')
    plt.plot(num_epochs, val_acc, 'b', label='Validation accuracy')
    plt.xlabel("numbers of epochs")
    plt.ylabel("accuracy")
    plt.grid(linestyle='--')
    plt.title('Training and validation accuracy')
    plt.legend()

    #------------------------------------------------
    # Plot training and validation loss per epoch
    #------------------------------------------------
    plt.subplot(1, 2, 2)
    plt.plot(num_epochs, train_loss, 'bo', label='Training loss')
    plt.plot(num_epochs, val_loss, 'b', label='Validation loss')
    plt.xlabel("numbers of epochs")
    plt.ylabel("loss")
    plt.grid(linestyle='--')
    plt.title('Training and validation loss')
    plt.legend()

    plt.show()

graph_plot(history)

Task #7 : Data Augmentation and Re-training of the model

• Use Data Augmentation to improve the perfromances of the model and reduce overfitting
• Re-training of the model
• Visualize the impact of Data Augmentation

# Create run-time augmentation on training and test dataset
# For training datagenerator, we add normalization and validation_split
train_datagen = ImageDataGenerator(rescale = 1 / 255.0,
                                   rotation_range = 45,
                                   width_shift_range = 0.3,
                                   height_shift_range = 0.3,
                                   shear_range = 0.4,
                                   zoom_range = 0.4,
                                   horizontal_flip = True,
                                   fill_mode = 'nearest',
                                   validation_split = 0.1112)

# For test datagenerator, we only normalize the data.
test_datagen = ImageDataGenerator(rescale=1 / 255.0)

# Flow training images in batches of 128 using train_datagen generator
training_generator = train_datagen.flow_from_directory(
    TRAINING_DIR,                # This is the source directory for training images
    target_size = (150, 150), # All images will be resized to 300x300
    batch_size = 20,
    subset = 'training',
    class_mode = 'binary')    # Since we use binary_crossentropy loss, we need binary labels

# Flow training images in batches of 128 using train_datagen generator
validation_generator = train_datagen.flow_from_directory(
        TRAINING_DIR,  # This is the source directory for training images
        target_size = (150, 150),           # All images will be resized to 150x150
        batch_size = 20,
        subset = 'validation',
        class_mode ='binary')  # Since we use binary_crossentropy loss, we need binary labels

# Flow training images in batches of 128 using train_datagen generator
testing_generator = test_datagen.flow_from_directory(
        TESTING_DIR,  # This is the source directory for training images
        target_size = (150, 150),           # All images will be resized to 150x150
        batch_size = 20,
        class_mode ='binary')  # Since we use binary_crossentropy loss, we need binary labels

Found 2400 images belonging to 2 classes.
Found 300 images belonging to 2 classes.
Found 300 images belonging to 2 classes.

history2 = model.fit(
    training_generator,
    steps_per_epoch = training_generator.n // 20,
    epochs = 20,
    validation_data = validation_generator,
    validation_steps = validation_generator.n // 20,
    callbacks = [callbacks, earlystopping],
    verbose = 2
)

Epoch 1/20
120/120 - 67s - loss: 0.3932 - accuracy: 0.8242 - val_loss: 0.3783 - val_accuracy: 0.8467
Epoch 2/20
120/120 - 69s - loss: 0.3696 - accuracy: 0.8354 - val_loss: 0.3648 - val_accuracy: 0.8433
Epoch 3/20
120/120 - 69s - loss: 0.3998 - accuracy: 0.8200 - val_loss: 0.4152 - val_accuracy: 0.8300
Epoch 4/20
120/120 - 71s - loss: 0.3716 - accuracy: 0.8288 - val_loss: 0.3748 - val_accuracy: 0.8500
Epoch 5/20
120/120 - 67s - loss: 0.3964 - accuracy: 0.8325 - val_loss: 0.4517 - val_accuracy: 0.8100
Epoch 6/20
120/120 - 70s - loss: 0.3796 - accuracy: 0.8313 - val_loss: 0.3790 - val_accuracy: 0.8400
Epoch 7/20
120/120 - 67s - loss: 0.3723 - accuracy: 0.8388 - val_loss: 0.4969 - val_accuracy: 0.7833
Epoch 8/20
120/120 - 72s - loss: 0.3803 - accuracy: 0.8304 - val_loss: 0.3611 - val_accuracy: 0.8400
Epoch 9/20
120/120 - 70s - loss: 0.3778 - accuracy: 0.8346 - val_loss: 0.4146 - val_accuracy: 0.8167
Epoch 10/20
120/120 - 68s - loss: 0.3762 - accuracy: 0.8375 - val_loss: 0.5552 - val_accuracy: 0.7533
Epoch 11/20
120/120 - 70s - loss: 0.3713 - accuracy: 0.8404 - val_loss: 0.3857 - val_accuracy: 0.8300
Epoch 12/20
120/120 - 67s - loss: 0.3950 - accuracy: 0.8221 - val_loss: 0.4406 - val_accuracy: 0.8267
Epoch 13/20
120/120 - 71s - loss: 0.3896 - accuracy: 0.8200 - val_loss: 0.3851 - val_accuracy: 0.8333
Epoch 14/20
120/120 - 68s - loss: 0.3832 - accuracy: 0.8342 - val_loss: 0.4732 - val_accuracy: 0.7700
Epoch 15/20
120/120 - 69s - loss: 0.3837 - accuracy: 0.8313 - val_loss: 0.4318 - val_accuracy: 0.8000
Epoch 16/20
120/120 - 70s - loss: 0.3830 - accuracy: 0.8329 - val_loss: 0.4189 - val_accuracy: 0.8100
Epoch 17/20
120/120 - 68s - loss: 0.3742 - accuracy: 0.8300 - val_loss: 0.4369 - val_accuracy: 0.8067
Epoch 18/20
120/120 - 71s - loss: 0.3796 - accuracy: 0.8379 - val_loss: 0.3325 - val_accuracy: 0.8533
Epoch 19/20
120/120 - 67s - loss: 0.3774 - accuracy: 0.8350 - val_loss: 0.4061 - val_accuracy: 0.8233
Epoch 20/20
120/120 - 71s - loss: 0.4035 - accuracy: 0.8225 - val_loss: 0.4578 - val_accuracy: 0.7867

graph_plot(history2)

Task #8 : Transfert Learning with Data Augmentation

Use transfert learning to improve performances of the model. We use the weights of a pre-trained InceptionV3 model on the ImageNet database for the transfer learning

• Import a model inception architecture and load the weigths file
• Choice of a layer in the InceptionV3 Model
• Building of the Model
• Trained model
• Visualization of the Accuracy and Loss for the Model

Import a model inception architecture and load the weigths file

from tensorflow.keras import layers, Model

# Import the inception model 
from tensorflow.keras.applications.inception_v3 import InceptionV3

weights_file = "./inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5"

# Create an instance of the inception model
pre_trained_model = InceptionV3(
    include_top = False,
    input_shape = (150, 150, 3),
    weights = None
)

# Loading weights
pre_trained_model.load_weights(weights_file)

# Summary of the model
pre_trained_model.summary()

# Make all the layers in the pre-trained model non-trainable
for layer in pre_trained_model.layers:
    layer.trainable = True

Model: "inception_v3"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_3 (InputLayer)            [(None, 150, 150, 3) 0                                            
__________________________________________________________________________________________________
conv2d_98 (Conv2D)              (None, 74, 74, 32)   864         input_3[0][0]                    
__________________________________________________________________________________________________
batch_normalization_94 (BatchNo (None, 74, 74, 32)   96          conv2d_98[0][0]                  
__________________________________________________________________________________________________
activation_94 (Activation)      (None, 74, 74, 32)   0           batch_normalization_94[0][0]     
__________________________________________________________________________________________________
conv2d_99 (Conv2D)              (None, 72, 72, 32)   9216        activation_94[0][0]              
__________________________________________________________________________________________________
batch_normalization_95 (BatchNo (None, 72, 72, 32)   96          conv2d_99[0][0]                  
__________________________________________________________________________________________________
activation_95 (Activation)      (None, 72, 72, 32)   0           batch_normalization_95[0][0]     
__________________________________________________________________________________________________
conv2d_100 (Conv2D)             (None, 72, 72, 64)   18432       activation_95[0][0]              
__________________________________________________________________________________________________
batch_normalization_96 (BatchNo (None, 72, 72, 64)   192         conv2d_100[0][0]                 
__________________________________________________________________________________________________
activation_96 (Activation)      (None, 72, 72, 64)   0           batch_normalization_96[0][0]     
__________________________________________________________________________________________________
max_pooling2d_8 (MaxPooling2D)  (None, 35, 35, 64)   0           activation_96[0][0]              
__________________________________________________________________________________________________
conv2d_101 (Conv2D)             (None, 35, 35, 80)   5120        max_pooling2d_8[0][0]            
__________________________________________________________________________________________________
batch_normalization_97 (BatchNo (None, 35, 35, 80)   240         conv2d_101[0][0]                 
__________________________________________________________________________________________________
activation_97 (Activation)      (None, 35, 35, 80)   0           batch_normalization_97[0][0]     
__________________________________________________________________________________________________
conv2d_102 (Conv2D)             (None, 33, 33, 192)  138240      activation_97[0][0]              
__________________________________________________________________________________________________
batch_normalization_98 (BatchNo (None, 33, 33, 192)  576         conv2d_102[0][0]                 
__________________________________________________________________________________________________
activation_98 (Activation)      (None, 33, 33, 192)  0           batch_normalization_98[0][0]     
__________________________________________________________________________________________________
max_pooling2d_9 (MaxPooling2D)  (None, 16, 16, 192)  0           activation_98[0][0]              
__________________________________________________________________________________________________
conv2d_106 (Conv2D)             (None, 16, 16, 64)   12288       max_pooling2d_9[0][0]            
__________________________________________________________________________________________________
batch_normalization_102 (BatchN (None, 16, 16, 64)   192         conv2d_106[0][0]                 
__________________________________________________________________________________________________
activation_102 (Activation)     (None, 16, 16, 64)   0           batch_normalization_102[0][0]    
__________________________________________________________________________________________________
conv2d_104 (Conv2D)             (None, 16, 16, 48)   9216        max_pooling2d_9[0][0]            
__________________________________________________________________________________________________
conv2d_107 (Conv2D)             (None, 16, 16, 96)   55296       activation_102[0][0]             
__________________________________________________________________________________________________
batch_normalization_100 (BatchN (None, 16, 16, 48)   144         conv2d_104[0][0]                 
__________________________________________________________________________________________________
batch_normalization_103 (BatchN (None, 16, 16, 96)   288         conv2d_107[0][0]                 
__________________________________________________________________________________________________
activation_100 (Activation)     (None, 16, 16, 48)   0           batch_normalization_100[0][0]    
__________________________________________________________________________________________________
activation_103 (Activation)     (None, 16, 16, 96)   0           batch_normalization_103[0][0]    
__________________________________________________________________________________________________
average_pooling2d_9 (AveragePoo (None, 16, 16, 192)  0           max_pooling2d_9[0][0]            
__________________________________________________________________________________________________
conv2d_103 (Conv2D)             (None, 16, 16, 64)   12288       max_pooling2d_9[0][0]            
__________________________________________________________________________________________________
conv2d_105 (Conv2D)             (None, 16, 16, 64)   76800       activation_100[0][0]             
__________________________________________________________________________________________________
conv2d_108 (Conv2D)             (None, 16, 16, 96)   82944       activation_103[0][0]             
__________________________________________________________________________________________________
conv2d_109 (Conv2D)             (None, 16, 16, 32)   6144        average_pooling2d_9[0][0]        
__________________________________________________________________________________________________
batch_normalization_99 (BatchNo (None, 16, 16, 64)   192         conv2d_103[0][0]                 
__________________________________________________________________________________________________
batch_normalization_101 (BatchN (None, 16, 16, 64)   192         conv2d_105[0][0]                 
__________________________________________________________________________________________________
batch_normalization_104 (BatchN (None, 16, 16, 96)   288         conv2d_108[0][0]                 
__________________________________________________________________________________________________
batch_normalization_105 (BatchN (None, 16, 16, 32)   96          conv2d_109[0][0]                 
__________________________________________________________________________________________________
activation_99 (Activation)      (None, 16, 16, 64)   0           batch_normalization_99[0][0]     
__________________________________________________________________________________________________
activation_101 (Activation)     (None, 16, 16, 64)   0           batch_normalization_101[0][0]    
__________________________________________________________________________________________________
activation_104 (Activation)     (None, 16, 16, 96)   0           batch_normalization_104[0][0]    
__________________________________________________________________________________________________
activation_105 (Activation)     (None, 16, 16, 32)   0           batch_normalization_105[0][0]    
__________________________________________________________________________________________________
mixed0 (Concatenate)            (None, 16, 16, 256)  0           activation_99[0][0]              
                                                                 activation_101[0][0]             
                                                                 activation_104[0][0]             
                                                                 activation_105[0][0]             
__________________________________________________________________________________________________
conv2d_113 (Conv2D)             (None, 16, 16, 64)   16384       mixed0[0][0]                     
__________________________________________________________________________________________________
batch_normalization_109 (BatchN (None, 16, 16, 64)   192         conv2d_113[0][0]                 
__________________________________________________________________________________________________
activation_109 (Activation)     (None, 16, 16, 64)   0           batch_normalization_109[0][0]    
__________________________________________________________________________________________________
conv2d_111 (Conv2D)             (None, 16, 16, 48)   12288       mixed0[0][0]                     
__________________________________________________________________________________________________
conv2d_114 (Conv2D)             (None, 16, 16, 96)   55296       activation_109[0][0]             
__________________________________________________________________________________________________
batch_normalization_107 (BatchN (None, 16, 16, 48)   144         conv2d_111[0][0]                 
__________________________________________________________________________________________________
batch_normalization_110 (BatchN (None, 16, 16, 96)   288         conv2d_114[0][0]                 
__________________________________________________________________________________________________
activation_107 (Activation)     (None, 16, 16, 48)   0           batch_normalization_107[0][0]    
__________________________________________________________________________________________________
activation_110 (Activation)     (None, 16, 16, 96)   0           batch_normalization_110[0][0]    
__________________________________________________________________________________________________
average_pooling2d_10 (AveragePo (None, 16, 16, 256)  0           mixed0[0][0]                     
__________________________________________________________________________________________________
conv2d_110 (Conv2D)             (None, 16, 16, 64)   16384       mixed0[0][0]                     
__________________________________________________________________________________________________
conv2d_112 (Conv2D)             (None, 16, 16, 64)   76800       activation_107[0][0]             
__________________________________________________________________________________________________
conv2d_115 (Conv2D)             (None, 16, 16, 96)   82944       activation_110[0][0]             
__________________________________________________________________________________________________
conv2d_116 (Conv2D)             (None, 16, 16, 64)   16384       average_pooling2d_10[0][0]       
__________________________________________________________________________________________________
batch_normalization_106 (BatchN (None, 16, 16, 64)   192         conv2d_110[0][0]                 
__________________________________________________________________________________________________
batch_normalization_108 (BatchN (None, 16, 16, 64)   192         conv2d_112[0][0]                 
__________________________________________________________________________________________________
batch_normalization_111 (BatchN (None, 16, 16, 96)   288         conv2d_115[0][0]                 
__________________________________________________________________________________________________
batch_normalization_112 (BatchN (None, 16, 16, 64)   192         conv2d_116[0][0]                 
__________________________________________________________________________________________________
activation_106 (Activation)     (None, 16, 16, 64)   0           batch_normalization_106[0][0]    
__________________________________________________________________________________________________
activation_108 (Activation)     (None, 16, 16, 64)   0           batch_normalization_108[0][0]    
__________________________________________________________________________________________________
activation_111 (Activation)     (None, 16, 16, 96)   0           batch_normalization_111[0][0]    
__________________________________________________________________________________________________
activation_112 (Activation)     (None, 16, 16, 64)   0           batch_normalization_112[0][0]    
__________________________________________________________________________________________________
mixed1 (Concatenate)            (None, 16, 16, 288)  0           activation_106[0][0]             
                                                                 activation_108[0][0]             
                                                                 activation_111[0][0]             
                                                                 activation_112[0][0]             
__________________________________________________________________________________________________
conv2d_120 (Conv2D)             (None, 16, 16, 64)   18432       mixed1[0][0]                     
__________________________________________________________________________________________________
batch_normalization_116 (BatchN (None, 16, 16, 64)   192         conv2d_120[0][0]                 
__________________________________________________________________________________________________
activation_116 (Activation)     (None, 16, 16, 64)   0           batch_normalization_116[0][0]    
__________________________________________________________________________________________________
conv2d_118 (Conv2D)             (None, 16, 16, 48)   13824       mixed1[0][0]                     
__________________________________________________________________________________________________
conv2d_121 (Conv2D)             (None, 16, 16, 96)   55296       activation_116[0][0]             
__________________________________________________________________________________________________
batch_normalization_114 (BatchN (None, 16, 16, 48)   144         conv2d_118[0][0]                 
__________________________________________________________________________________________________
batch_normalization_117 (BatchN (None, 16, 16, 96)   288         conv2d_121[0][0]                 
__________________________________________________________________________________________________
activation_114 (Activation)     (None, 16, 16, 48)   0           batch_normalization_114[0][0]    
__________________________________________________________________________________________________
activation_117 (Activation)     (None, 16, 16, 96)   0           batch_normalization_117[0][0]    
__________________________________________________________________________________________________
average_pooling2d_11 (AveragePo (None, 16, 16, 288)  0           mixed1[0][0]                     
__________________________________________________________________________________________________
conv2d_117 (Conv2D)             (None, 16, 16, 64)   18432       mixed1[0][0]                     
__________________________________________________________________________________________________
conv2d_119 (Conv2D)             (None, 16, 16, 64)   76800       activation_114[0][0]             
__________________________________________________________________________________________________
conv2d_122 (Conv2D)             (None, 16, 16, 96)   82944       activation_117[0][0]             
__________________________________________________________________________________________________
conv2d_123 (Conv2D)             (None, 16, 16, 64)   18432       average_pooling2d_11[0][0]       
__________________________________________________________________________________________________
batch_normalization_113 (BatchN (None, 16, 16, 64)   192         conv2d_117[0][0]                 
__________________________________________________________________________________________________
batch_normalization_115 (BatchN (None, 16, 16, 64)   192         conv2d_119[0][0]                 
__________________________________________________________________________________________________
batch_normalization_118 (BatchN (None, 16, 16, 96)   288         conv2d_122[0][0]                 
__________________________________________________________________________________________________
batch_normalization_119 (BatchN (None, 16, 16, 64)   192         conv2d_123[0][0]                 
__________________________________________________________________________________________________
activation_113 (Activation)     (None, 16, 16, 64)   0           batch_normalization_113[0][0]    
__________________________________________________________________________________________________
activation_115 (Activation)     (None, 16, 16, 64)   0           batch_normalization_115[0][0]    
__________________________________________________________________________________________________
activation_118 (Activation)     (None, 16, 16, 96)   0           batch_normalization_118[0][0]    
__________________________________________________________________________________________________
activation_119 (Activation)     (None, 16, 16, 64)   0           batch_normalization_119[0][0]    
__________________________________________________________________________________________________
mixed2 (Concatenate)            (None, 16, 16, 288)  0           activation_113[0][0]             
                                                                 activation_115[0][0]             
                                                                 activation_118[0][0]             
                                                                 activation_119[0][0]             
__________________________________________________________________________________________________
conv2d_125 (Conv2D)             (None, 16, 16, 64)   18432       mixed2[0][0]                     
__________________________________________________________________________________________________
batch_normalization_121 (BatchN (None, 16, 16, 64)   192         conv2d_125[0][0]                 
__________________________________________________________________________________________________
activation_121 (Activation)     (None, 16, 16, 64)   0           batch_normalization_121[0][0]    
__________________________________________________________________________________________________
conv2d_126 (Conv2D)             (None, 16, 16, 96)   55296       activation_121[0][0]             
__________________________________________________________________________________________________
batch_normalization_122 (BatchN (None, 16, 16, 96)   288         conv2d_126[0][0]                 
__________________________________________________________________________________________________
activation_122 (Activation)     (None, 16, 16, 96)   0           batch_normalization_122[0][0]    
__________________________________________________________________________________________________
conv2d_124 (Conv2D)             (None, 7, 7, 384)    995328      mixed2[0][0]                     
__________________________________________________________________________________________________
conv2d_127 (Conv2D)             (None, 7, 7, 96)     82944       activation_122[0][0]             
__________________________________________________________________________________________________
batch_normalization_120 (BatchN (None, 7, 7, 384)    1152        conv2d_124[0][0]                 
__________________________________________________________________________________________________
batch_normalization_123 (BatchN (None, 7, 7, 96)     288         conv2d_127[0][0]                 
__________________________________________________________________________________________________
activation_120 (Activation)     (None, 7, 7, 384)    0           batch_normalization_120[0][0]    
__________________________________________________________________________________________________
activation_123 (Activation)     (None, 7, 7, 96)     0           batch_normalization_123[0][0]    
__________________________________________________________________________________________________
max_pooling2d_10 (MaxPooling2D) (None, 7, 7, 288)    0           mixed2[0][0]                     
__________________________________________________________________________________________________
mixed3 (Concatenate)            (None, 7, 7, 768)    0           activation_120[0][0]             
                                                                 activation_123[0][0]             
                                                                 max_pooling2d_10[0][0]           
__________________________________________________________________________________________________
conv2d_132 (Conv2D)             (None, 7, 7, 128)    98304       mixed3[0][0]                     
__________________________________________________________________________________________________
batch_normalization_128 (BatchN (None, 7, 7, 128)    384         conv2d_132[0][0]                 
__________________________________________________________________________________________________
activation_128 (Activation)     (None, 7, 7, 128)    0           batch_normalization_128[0][0]    
__________________________________________________________________________________________________
conv2d_133 (Conv2D)             (None, 7, 7, 128)    114688      activation_128[0][0]             
__________________________________________________________________________________________________
batch_normalization_129 (BatchN (None, 7, 7, 128)    384         conv2d_133[0][0]                 
__________________________________________________________________________________________________
activation_129 (Activation)     (None, 7, 7, 128)    0           batch_normalization_129[0][0]    
__________________________________________________________________________________________________
conv2d_129 (Conv2D)             (None, 7, 7, 128)    98304       mixed3[0][0]                     
__________________________________________________________________________________________________
conv2d_134 (Conv2D)             (None, 7, 7, 128)    114688      activation_129[0][0]             
__________________________________________________________________________________________________
batch_normalization_125 (BatchN (None, 7, 7, 128)    384         conv2d_129[0][0]                 
__________________________________________________________________________________________________
batch_normalization_130 (BatchN (None, 7, 7, 128)    384         conv2d_134[0][0]                 
__________________________________________________________________________________________________
activation_125 (Activation)     (None, 7, 7, 128)    0           batch_normalization_125[0][0]    
__________________________________________________________________________________________________
activation_130 (Activation)     (None, 7, 7, 128)    0           batch_normalization_130[0][0]    
__________________________________________________________________________________________________
conv2d_130 (Conv2D)             (None, 7, 7, 128)    114688      activation_125[0][0]             
__________________________________________________________________________________________________
conv2d_135 (Conv2D)             (None, 7, 7, 128)    114688      activation_130[0][0]             
__________________________________________________________________________________________________
batch_normalization_126 (BatchN (None, 7, 7, 128)    384         conv2d_130[0][0]                 
__________________________________________________________________________________________________
batch_normalization_131 (BatchN (None, 7, 7, 128)    384         conv2d_135[0][0]                 
__________________________________________________________________________________________________
activation_126 (Activation)     (None, 7, 7, 128)    0           batch_normalization_126[0][0]    
__________________________________________________________________________________________________
activation_131 (Activation)     (None, 7, 7, 128)    0           batch_normalization_131[0][0]    
__________________________________________________________________________________________________
average_pooling2d_12 (AveragePo (None, 7, 7, 768)    0           mixed3[0][0]                     
__________________________________________________________________________________________________
conv2d_128 (Conv2D)             (None, 7, 7, 192)    147456      mixed3[0][0]                     
__________________________________________________________________________________________________
conv2d_131 (Conv2D)             (None, 7, 7, 192)    172032      activation_126[0][0]             
__________________________________________________________________________________________________
conv2d_136 (Conv2D)             (None, 7, 7, 192)    172032      activation_131[0][0]             
__________________________________________________________________________________________________
conv2d_137 (Conv2D)             (None, 7, 7, 192)    147456      average_pooling2d_12[0][0]       
__________________________________________________________________________________________________
batch_normalization_124 (BatchN (None, 7, 7, 192)    576         conv2d_128[0][0]                 
__________________________________________________________________________________________________
batch_normalization_127 (BatchN (None, 7, 7, 192)    576         conv2d_131[0][0]                 
__________________________________________________________________________________________________
batch_normalization_132 (BatchN (None, 7, 7, 192)    576         conv2d_136[0][0]                 
__________________________________________________________________________________________________
batch_normalization_133 (BatchN (None, 7, 7, 192)    576         conv2d_137[0][0]                 
__________________________________________________________________________________________________
activation_124 (Activation)     (None, 7, 7, 192)    0           batch_normalization_124[0][0]    
__________________________________________________________________________________________________
activation_127 (Activation)     (None, 7, 7, 192)    0           batch_normalization_127[0][0]    
__________________________________________________________________________________________________
activation_132 (Activation)     (None, 7, 7, 192)    0           batch_normalization_132[0][0]    
__________________________________________________________________________________________________
activation_133 (Activation)     (None, 7, 7, 192)    0           batch_normalization_133[0][0]    
__________________________________________________________________________________________________
mixed4 (Concatenate)            (None, 7, 7, 768)    0           activation_124[0][0]             
                                                                 activation_127[0][0]             
                                                                 activation_132[0][0]             
                                                                 activation_133[0][0]             
__________________________________________________________________________________________________
conv2d_142 (Conv2D)             (None, 7, 7, 160)    122880      mixed4[0][0]                     
__________________________________________________________________________________________________
batch_normalization_138 (BatchN (None, 7, 7, 160)    480         conv2d_142[0][0]                 
__________________________________________________________________________________________________
activation_138 (Activation)     (None, 7, 7, 160)    0           batch_normalization_138[0][0]    
__________________________________________________________________________________________________
conv2d_143 (Conv2D)             (None, 7, 7, 160)    179200      activation_138[0][0]             
__________________________________________________________________________________________________
batch_normalization_139 (BatchN (None, 7, 7, 160)    480         conv2d_143[0][0]                 
__________________________________________________________________________________________________
activation_139 (Activation)     (None, 7, 7, 160)    0           batch_normalization_139[0][0]    
__________________________________________________________________________________________________
conv2d_139 (Conv2D)             (None, 7, 7, 160)    122880      mixed4[0][0]                     
__________________________________________________________________________________________________
conv2d_144 (Conv2D)             (None, 7, 7, 160)    179200      activation_139[0][0]             
__________________________________________________________________________________________________
batch_normalization_135 (BatchN (None, 7, 7, 160)    480         conv2d_139[0][0]                 
__________________________________________________________________________________________________
batch_normalization_140 (BatchN (None, 7, 7, 160)    480         conv2d_144[0][0]                 
__________________________________________________________________________________________________
activation_135 (Activation)     (None, 7, 7, 160)    0           batch_normalization_135[0][0]    
__________________________________________________________________________________________________
activation_140 (Activation)     (None, 7, 7, 160)    0           batch_normalization_140[0][0]    
__________________________________________________________________________________________________
conv2d_140 (Conv2D)             (None, 7, 7, 160)    179200      activation_135[0][0]             
__________________________________________________________________________________________________
conv2d_145 (Conv2D)             (None, 7, 7, 160)    179200      activation_140[0][0]             
__________________________________________________________________________________________________
batch_normalization_136 (BatchN (None, 7, 7, 160)    480         conv2d_140[0][0]                 
__________________________________________________________________________________________________
batch_normalization_141 (BatchN (None, 7, 7, 160)    480         conv2d_145[0][0]                 
__________________________________________________________________________________________________
activation_136 (Activation)     (None, 7, 7, 160)    0           batch_normalization_136[0][0]    
__________________________________________________________________________________________________
activation_141 (Activation)     (None, 7, 7, 160)    0           batch_normalization_141[0][0]    
__________________________________________________________________________________________________
average_pooling2d_13 (AveragePo (None, 7, 7, 768)    0           mixed4[0][0]                     
__________________________________________________________________________________________________
conv2d_138 (Conv2D)             (None, 7, 7, 192)    147456      mixed4[0][0]                     
__________________________________________________________________________________________________
conv2d_141 (Conv2D)             (None, 7, 7, 192)    215040      activation_136[0][0]             
__________________________________________________________________________________________________
conv2d_146 (Conv2D)             (None, 7, 7, 192)    215040      activation_141[0][0]             
__________________________________________________________________________________________________
conv2d_147 (Conv2D)             (None, 7, 7, 192)    147456      average_pooling2d_13[0][0]       
__________________________________________________________________________________________________
batch_normalization_134 (BatchN (None, 7, 7, 192)    576         conv2d_138[0][0]                 
__________________________________________________________________________________________________
batch_normalization_137 (BatchN (None, 7, 7, 192)    576         conv2d_141[0][0]                 
__________________________________________________________________________________________________
batch_normalization_142 (BatchN (None, 7, 7, 192)    576         conv2d_146[0][0]                 
__________________________________________________________________________________________________
batch_normalization_143 (BatchN (None, 7, 7, 192)    576         conv2d_147[0][0]                 
__________________________________________________________________________________________________
activation_134 (Activation)     (None, 7, 7, 192)    0           batch_normalization_134[0][0]    
__________________________________________________________________________________________________
activation_137 (Activation)     (None, 7, 7, 192)    0           batch_normalization_137[0][0]    
__________________________________________________________________________________________________
activation_142 (Activation)     (None, 7, 7, 192)    0           batch_normalization_142[0][0]    
__________________________________________________________________________________________________
activation_143 (Activation)     (None, 7, 7, 192)    0           batch_normalization_143[0][0]    
__________________________________________________________________________________________________
mixed5 (Concatenate)            (None, 7, 7, 768)    0           activation_134[0][0]             
                                                                 activation_137[0][0]             
                                                                 activation_142[0][0]             
                                                                 activation_143[0][0]             
__________________________________________________________________________________________________
conv2d_152 (Conv2D)             (None, 7, 7, 160)    122880      mixed5[0][0]                     
__________________________________________________________________________________________________
batch_normalization_148 (BatchN (None, 7, 7, 160)    480         conv2d_152[0][0]                 
__________________________________________________________________________________________________
activation_148 (Activation)     (None, 7, 7, 160)    0           batch_normalization_148[0][0]    
__________________________________________________________________________________________________
conv2d_153 (Conv2D)             (None, 7, 7, 160)    179200      activation_148[0][0]             
__________________________________________________________________________________________________
batch_normalization_149 (BatchN (None, 7, 7, 160)    480         conv2d_153[0][0]                 
__________________________________________________________________________________________________
activation_149 (Activation)     (None, 7, 7, 160)    0           batch_normalization_149[0][0]    
__________________________________________________________________________________________________
conv2d_149 (Conv2D)             (None, 7, 7, 160)    122880      mixed5[0][0]                     
__________________________________________________________________________________________________
conv2d_154 (Conv2D)             (None, 7, 7, 160)    179200      activation_149[0][0]             
__________________________________________________________________________________________________
batch_normalization_145 (BatchN (None, 7, 7, 160)    480         conv2d_149[0][0]                 
__________________________________________________________________________________________________
batch_normalization_150 (BatchN (None, 7, 7, 160)    480         conv2d_154[0][0]                 
__________________________________________________________________________________________________
activation_145 (Activation)     (None, 7, 7, 160)    0           batch_normalization_145[0][0]    
__________________________________________________________________________________________________
activation_150 (Activation)     (None, 7, 7, 160)    0           batch_normalization_150[0][0]    
__________________________________________________________________________________________________
conv2d_150 (Conv2D)             (None, 7, 7, 160)    179200      activation_145[0][0]             
__________________________________________________________________________________________________
conv2d_155 (Conv2D)             (None, 7, 7, 160)    179200      activation_150[0][0]             
__________________________________________________________________________________________________
batch_normalization_146 (BatchN (None, 7, 7, 160)    480         conv2d_150[0][0]                 
__________________________________________________________________________________________________
batch_normalization_151 (BatchN (None, 7, 7, 160)    480         conv2d_155[0][0]                 
__________________________________________________________________________________________________
activation_146 (Activation)     (None, 7, 7, 160)    0           batch_normalization_146[0][0]    
__________________________________________________________________________________________________
activation_151 (Activation)     (None, 7, 7, 160)    0           batch_normalization_151[0][0]    
__________________________________________________________________________________________________
average_pooling2d_14 (AveragePo (None, 7, 7, 768)    0           mixed5[0][0]                     
__________________________________________________________________________________________________
conv2d_148 (Conv2D)             (None, 7, 7, 192)    147456      mixed5[0][0]                     
__________________________________________________________________________________________________
conv2d_151 (Conv2D)             (None, 7, 7, 192)    215040      activation_146[0][0]             
__________________________________________________________________________________________________
conv2d_156 (Conv2D)             (None, 7, 7, 192)    215040      activation_151[0][0]             
__________________________________________________________________________________________________
conv2d_157 (Conv2D)             (None, 7, 7, 192)    147456      average_pooling2d_14[0][0]       
__________________________________________________________________________________________________
batch_normalization_144 (BatchN (None, 7, 7, 192)    576         conv2d_148[0][0]                 
__________________________________________________________________________________________________
batch_normalization_147 (BatchN (None, 7, 7, 192)    576         conv2d_151[0][0]                 
__________________________________________________________________________________________________
batch_normalization_152 (BatchN (None, 7, 7, 192)    576         conv2d_156[0][0]                 
__________________________________________________________________________________________________
batch_normalization_153 (BatchN (None, 7, 7, 192)    576         conv2d_157[0][0]                 
__________________________________________________________________________________________________
activation_144 (Activation)     (None, 7, 7, 192)    0           batch_normalization_144[0][0]    
__________________________________________________________________________________________________
activation_147 (Activation)     (None, 7, 7, 192)    0           batch_normalization_147[0][0]    
__________________________________________________________________________________________________
activation_152 (Activation)     (None, 7, 7, 192)    0           batch_normalization_152[0][0]    
__________________________________________________________________________________________________
activation_153 (Activation)     (None, 7, 7, 192)    0           batch_normalization_153[0][0]    
__________________________________________________________________________________________________
mixed6 (Concatenate)            (None, 7, 7, 768)    0           activation_144[0][0]             
                                                                 activation_147[0][0]             
                                                                 activation_152[0][0]             
                                                                 activation_153[0][0]             
__________________________________________________________________________________________________
conv2d_162 (Conv2D)             (None, 7, 7, 192)    147456      mixed6[0][0]                     
__________________________________________________________________________________________________
batch_normalization_158 (BatchN (None, 7, 7, 192)    576         conv2d_162[0][0]                 
__________________________________________________________________________________________________
activation_158 (Activation)     (None, 7, 7, 192)    0           batch_normalization_158[0][0]    
__________________________________________________________________________________________________
conv2d_163 (Conv2D)             (None, 7, 7, 192)    258048      activation_158[0][0]             
__________________________________________________________________________________________________
batch_normalization_159 (BatchN (None, 7, 7, 192)    576         conv2d_163[0][0]                 
__________________________________________________________________________________________________
activation_159 (Activation)     (None, 7, 7, 192)    0           batch_normalization_159[0][0]    
__________________________________________________________________________________________________
conv2d_159 (Conv2D)             (None, 7, 7, 192)    147456      mixed6[0][0]                     
__________________________________________________________________________________________________
conv2d_164 (Conv2D)             (None, 7, 7, 192)    258048      activation_159[0][0]             
__________________________________________________________________________________________________
batch_normalization_155 (BatchN (None, 7, 7, 192)    576         conv2d_159[0][0]                 
__________________________________________________________________________________________________
batch_normalization_160 (BatchN (None, 7, 7, 192)    576         conv2d_164[0][0]                 
__________________________________________________________________________________________________
activation_155 (Activation)     (None, 7, 7, 192)    0           batch_normalization_155[0][0]    
__________________________________________________________________________________________________
activation_160 (Activation)     (None, 7, 7, 192)    0           batch_normalization_160[0][0]    
__________________________________________________________________________________________________
conv2d_160 (Conv2D)             (None, 7, 7, 192)    258048      activation_155[0][0]             
__________________________________________________________________________________________________
conv2d_165 (Conv2D)             (None, 7, 7, 192)    258048      activation_160[0][0]             
__________________________________________________________________________________________________
batch_normalization_156 (BatchN (None, 7, 7, 192)    576         conv2d_160[0][0]                 
__________________________________________________________________________________________________
batch_normalization_161 (BatchN (None, 7, 7, 192)    576         conv2d_165[0][0]                 
__________________________________________________________________________________________________
activation_156 (Activation)     (None, 7, 7, 192)    0           batch_normalization_156[0][0]    
__________________________________________________________________________________________________
activation_161 (Activation)     (None, 7, 7, 192)    0           batch_normalization_161[0][0]    
__________________________________________________________________________________________________
average_pooling2d_15 (AveragePo (None, 7, 7, 768)    0           mixed6[0][0]                     
__________________________________________________________________________________________________
conv2d_158 (Conv2D)             (None, 7, 7, 192)    147456      mixed6[0][0]                     
__________________________________________________________________________________________________
conv2d_161 (Conv2D)             (None, 7, 7, 192)    258048      activation_156[0][0]             
__________________________________________________________________________________________________
conv2d_166 (Conv2D)             (None, 7, 7, 192)    258048      activation_161[0][0]             
__________________________________________________________________________________________________
conv2d_167 (Conv2D)             (None, 7, 7, 192)    147456      average_pooling2d_15[0][0]       
__________________________________________________________________________________________________
batch_normalization_154 (BatchN (None, 7, 7, 192)    576         conv2d_158[0][0]                 
__________________________________________________________________________________________________
batch_normalization_157 (BatchN (None, 7, 7, 192)    576         conv2d_161[0][0]                 
__________________________________________________________________________________________________
batch_normalization_162 (BatchN (None, 7, 7, 192)    576         conv2d_166[0][0]                 
__________________________________________________________________________________________________
batch_normalization_163 (BatchN (None, 7, 7, 192)    576         conv2d_167[0][0]                 
__________________________________________________________________________________________________
activation_154 (Activation)     (None, 7, 7, 192)    0           batch_normalization_154[0][0]    
__________________________________________________________________________________________________
activation_157 (Activation)     (None, 7, 7, 192)    0           batch_normalization_157[0][0]    
__________________________________________________________________________________________________
activation_162 (Activation)     (None, 7, 7, 192)    0           batch_normalization_162[0][0]    
__________________________________________________________________________________________________
activation_163 (Activation)     (None, 7, 7, 192)    0           batch_normalization_163[0][0]    
__________________________________________________________________________________________________
mixed7 (Concatenate)            (None, 7, 7, 768)    0           activation_154[0][0]             
                                                                 activation_157[0][0]             
                                                                 activation_162[0][0]             
                                                                 activation_163[0][0]             
__________________________________________________________________________________________________
conv2d_170 (Conv2D)             (None, 7, 7, 192)    147456      mixed7[0][0]                     
__________________________________________________________________________________________________
batch_normalization_166 (BatchN (None, 7, 7, 192)    576         conv2d_170[0][0]                 
__________________________________________________________________________________________________
activation_166 (Activation)     (None, 7, 7, 192)    0           batch_normalization_166[0][0]    
__________________________________________________________________________________________________
conv2d_171 (Conv2D)             (None, 7, 7, 192)    258048      activation_166[0][0]             
__________________________________________________________________________________________________
batch_normalization_167 (BatchN (None, 7, 7, 192)    576         conv2d_171[0][0]                 
__________________________________________________________________________________________________
activation_167 (Activation)     (None, 7, 7, 192)    0           batch_normalization_167[0][0]    
__________________________________________________________________________________________________
conv2d_168 (Conv2D)             (None, 7, 7, 192)    147456      mixed7[0][0]                     
__________________________________________________________________________________________________
conv2d_172 (Conv2D)             (None, 7, 7, 192)    258048      activation_167[0][0]             
__________________________________________________________________________________________________
batch_normalization_164 (BatchN (None, 7, 7, 192)    576         conv2d_168[0][0]                 
__________________________________________________________________________________________________
batch_normalization_168 (BatchN (None, 7, 7, 192)    576         conv2d_172[0][0]                 
__________________________________________________________________________________________________
activation_164 (Activation)     (None, 7, 7, 192)    0           batch_normalization_164[0][0]    
__________________________________________________________________________________________________
activation_168 (Activation)     (None, 7, 7, 192)    0           batch_normalization_168[0][0]    
__________________________________________________________________________________________________
conv2d_169 (Conv2D)             (None, 3, 3, 320)    552960      activation_164[0][0]             
__________________________________________________________________________________________________
conv2d_173 (Conv2D)             (None, 3, 3, 192)    331776      activation_168[0][0]             
__________________________________________________________________________________________________
batch_normalization_165 (BatchN (None, 3, 3, 320)    960         conv2d_169[0][0]                 
__________________________________________________________________________________________________
batch_normalization_169 (BatchN (None, 3, 3, 192)    576         conv2d_173[0][0]                 
__________________________________________________________________________________________________
activation_165 (Activation)     (None, 3, 3, 320)    0           batch_normalization_165[0][0]    
__________________________________________________________________________________________________
activation_169 (Activation)     (None, 3, 3, 192)    0           batch_normalization_169[0][0]    
__________________________________________________________________________________________________
max_pooling2d_11 (MaxPooling2D) (None, 3, 3, 768)    0           mixed7[0][0]                     
__________________________________________________________________________________________________
mixed8 (Concatenate)            (None, 3, 3, 1280)   0           activation_165[0][0]             
                                                                 activation_169[0][0]             
                                                                 max_pooling2d_11[0][0]           
__________________________________________________________________________________________________
conv2d_178 (Conv2D)             (None, 3, 3, 448)    573440      mixed8[0][0]                     
__________________________________________________________________________________________________
batch_normalization_174 (BatchN (None, 3, 3, 448)    1344        conv2d_178[0][0]                 
__________________________________________________________________________________________________
activation_174 (Activation)     (None, 3, 3, 448)    0           batch_normalization_174[0][0]    
__________________________________________________________________________________________________
conv2d_175 (Conv2D)             (None, 3, 3, 384)    491520      mixed8[0][0]                     
__________________________________________________________________________________________________
conv2d_179 (Conv2D)             (None, 3, 3, 384)    1548288     activation_174[0][0]             
__________________________________________________________________________________________________
batch_normalization_171 (BatchN (None, 3, 3, 384)    1152        conv2d_175[0][0]                 
__________________________________________________________________________________________________
batch_normalization_175 (BatchN (None, 3, 3, 384)    1152        conv2d_179[0][0]                 
__________________________________________________________________________________________________
activation_171 (Activation)     (None, 3, 3, 384)    0           batch_normalization_171[0][0]    
__________________________________________________________________________________________________
activation_175 (Activation)     (None, 3, 3, 384)    0           batch_normalization_175[0][0]    
__________________________________________________________________________________________________
conv2d_176 (Conv2D)             (None, 3, 3, 384)    442368      activation_171[0][0]             
__________________________________________________________________________________________________
conv2d_177 (Conv2D)             (None, 3, 3, 384)    442368      activation_171[0][0]             
__________________________________________________________________________________________________
conv2d_180 (Conv2D)             (None, 3, 3, 384)    442368      activation_175[0][0]             
__________________________________________________________________________________________________
conv2d_181 (Conv2D)             (None, 3, 3, 384)    442368      activation_175[0][0]             
__________________________________________________________________________________________________
average_pooling2d_16 (AveragePo (None, 3, 3, 1280)   0           mixed8[0][0]                     
__________________________________________________________________________________________________
conv2d_174 (Conv2D)             (None, 3, 3, 320)    409600      mixed8[0][0]                     
__________________________________________________________________________________________________
batch_normalization_172 (BatchN (None, 3, 3, 384)    1152        conv2d_176[0][0]                 
__________________________________________________________________________________________________
batch_normalization_173 (BatchN (None, 3, 3, 384)    1152        conv2d_177[0][0]                 
__________________________________________________________________________________________________
batch_normalization_176 (BatchN (None, 3, 3, 384)    1152        conv2d_180[0][0]                 
__________________________________________________________________________________________________
batch_normalization_177 (BatchN (None, 3, 3, 384)    1152        conv2d_181[0][0]                 
__________________________________________________________________________________________________
conv2d_182 (Conv2D)             (None, 3, 3, 192)    245760      average_pooling2d_16[0][0]       
__________________________________________________________________________________________________
batch_normalization_170 (BatchN (None, 3, 3, 320)    960         conv2d_174[0][0]                 
__________________________________________________________________________________________________
activation_172 (Activation)     (None, 3, 3, 384)    0           batch_normalization_172[0][0]    
__________________________________________________________________________________________________
activation_173 (Activation)     (None, 3, 3, 384)    0           batch_normalization_173[0][0]    
__________________________________________________________________________________________________
activation_176 (Activation)     (None, 3, 3, 384)    0           batch_normalization_176[0][0]    
__________________________________________________________________________________________________
activation_177 (Activation)     (None, 3, 3, 384)    0           batch_normalization_177[0][0]    
__________________________________________________________________________________________________
batch_normalization_178 (BatchN (None, 3, 3, 192)    576         conv2d_182[0][0]                 
__________________________________________________________________________________________________
activation_170 (Activation)     (None, 3, 3, 320)    0           batch_normalization_170[0][0]    
__________________________________________________________________________________________________
mixed9_0 (Concatenate)          (None, 3, 3, 768)    0           activation_172[0][0]             
                                                                 activation_173[0][0]             
__________________________________________________________________________________________________
concatenate_2 (Concatenate)     (None, 3, 3, 768)    0           activation_176[0][0]             
                                                                 activation_177[0][0]             
__________________________________________________________________________________________________
activation_178 (Activation)     (None, 3, 3, 192)    0           batch_normalization_178[0][0]    
__________________________________________________________________________________________________
mixed9 (Concatenate)            (None, 3, 3, 2048)   0           activation_170[0][0]             
                                                                 mixed9_0[0][0]                   
                                                                 concatenate_2[0][0]              
                                                                 activation_178[0][0]             
__________________________________________________________________________________________________
conv2d_187 (Conv2D)             (None, 3, 3, 448)    917504      mixed9[0][0]                     
__________________________________________________________________________________________________
batch_normalization_183 (BatchN (None, 3, 3, 448)    1344        conv2d_187[0][0]                 
__________________________________________________________________________________________________
activation_183 (Activation)     (None, 3, 3, 448)    0           batch_normalization_183[0][0]    
__________________________________________________________________________________________________
conv2d_184 (Conv2D)             (None, 3, 3, 384)    786432      mixed9[0][0]                     
__________________________________________________________________________________________________
conv2d_188 (Conv2D)             (None, 3, 3, 384)    1548288     activation_183[0][0]             
__________________________________________________________________________________________________
batch_normalization_180 (BatchN (None, 3, 3, 384)    1152        conv2d_184[0][0]                 
__________________________________________________________________________________________________
batch_normalization_184 (BatchN (None, 3, 3, 384)    1152        conv2d_188[0][0]                 
__________________________________________________________________________________________________
activation_180 (Activation)     (None, 3, 3, 384)    0           batch_normalization_180[0][0]    
__________________________________________________________________________________________________
activation_184 (Activation)     (None, 3, 3, 384)    0           batch_normalization_184[0][0]    
__________________________________________________________________________________________________
conv2d_185 (Conv2D)             (None, 3, 3, 384)    442368      activation_180[0][0]             
__________________________________________________________________________________________________
conv2d_186 (Conv2D)             (None, 3, 3, 384)    442368      activation_180[0][0]             
__________________________________________________________________________________________________
conv2d_189 (Conv2D)             (None, 3, 3, 384)    442368      activation_184[0][0]             
__________________________________________________________________________________________________
conv2d_190 (Conv2D)             (None, 3, 3, 384)    442368      activation_184[0][0]             
__________________________________________________________________________________________________
average_pooling2d_17 (AveragePo (None, 3, 3, 2048)   0           mixed9[0][0]                     
__________________________________________________________________________________________________
conv2d_183 (Conv2D)             (None, 3, 3, 320)    655360      mixed9[0][0]                     
__________________________________________________________________________________________________
batch_normalization_181 (BatchN (None, 3, 3, 384)    1152        conv2d_185[0][0]                 
__________________________________________________________________________________________________
batch_normalization_182 (BatchN (None, 3, 3, 384)    1152        conv2d_186[0][0]                 
__________________________________________________________________________________________________
batch_normalization_185 (BatchN (None, 3, 3, 384)    1152        conv2d_189[0][0]                 
__________________________________________________________________________________________________
batch_normalization_186 (BatchN (None, 3, 3, 384)    1152        conv2d_190[0][0]                 
__________________________________________________________________________________________________
conv2d_191 (Conv2D)             (None, 3, 3, 192)    393216      average_pooling2d_17[0][0]       
__________________________________________________________________________________________________
batch_normalization_179 (BatchN (None, 3, 3, 320)    960         conv2d_183[0][0]                 
__________________________________________________________________________________________________
activation_181 (Activation)     (None, 3, 3, 384)    0           batch_normalization_181[0][0]    
__________________________________________________________________________________________________
activation_182 (Activation)     (None, 3, 3, 384)    0           batch_normalization_182[0][0]    
__________________________________________________________________________________________________
activation_185 (Activation)     (None, 3, 3, 384)    0           batch_normalization_185[0][0]    
__________________________________________________________________________________________________
activation_186 (Activation)     (None, 3, 3, 384)    0           batch_normalization_186[0][0]    
__________________________________________________________________________________________________
batch_normalization_187 (BatchN (None, 3, 3, 192)    576         conv2d_191[0][0]                 
__________________________________________________________________________________________________
activation_179 (Activation)     (None, 3, 3, 320)    0           batch_normalization_179[0][0]    
__________________________________________________________________________________________________
mixed9_1 (Concatenate)          (None, 3, 3, 768)    0           activation_181[0][0]             
                                                                 activation_182[0][0]             
__________________________________________________________________________________________________
concatenate_3 (Concatenate)     (None, 3, 3, 768)    0           activation_185[0][0]             
                                                                 activation_186[0][0]             
__________________________________________________________________________________________________
activation_187 (Activation)     (None, 3, 3, 192)    0           batch_normalization_187[0][0]    
__________________________________________________________________________________________________
mixed10 (Concatenate)           (None, 3, 3, 2048)   0           activation_179[0][0]             
                                                                 mixed9_1[0][0]                   
                                                                 concatenate_3[0][0]              
                                                                 activation_187[0][0]             
==================================================================================================
Total params: 21,802,784
Trainable params: 21,768,352
Non-trainable params: 34,432
__________________________________________________________________________________________________

Choice of a layer in the InceptionV3 Model

We select one layer : for example mixed8 layer.

# Select the layer
last_layer = pre_trained_model.get_layer('mixed8')

# Print the output shape of the layer
print('last layer output shape :',  last_layer.output_shape)

# Retrieve le output of the layer : tensor 4D
last_output = last_layer.output

last layer output shape : (None, 3, 3, 1280)

Building of the Model

We will use the Droupout layer to regularize the model

# Flatten the output layer to 1 dimension
x = tf.keras.layers.Flatten()(last_output)

# Add a fully connected layer with 1024 hidden units and ReLU activation
x = tf.keras.layers.Dense(units = 512, activation = 'relu')(x)

# Add a dropout rate of 0.5
x = tf.keras.layers.Dropout(rate = 0.5)(x)

# Add a final sigmoid layer for classification
outputs = tf.keras.layers.Dense(units = 1, activation = 'sigmoid')(x)

# Building of the model
model = Model(inputs = pre_trained_model.input, outputs = outputs)

# Compile the model
model.compile(optimizer=RMSprop(lr=1e-4), loss='binary_crossentropy', metrics = ['accuracy'])

# Summary
model.summary()

Model: "functional_11"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_3 (InputLayer)            [(None, 150, 150, 3) 0                                            
__________________________________________________________________________________________________
conv2d_98 (Conv2D)              (None, 74, 74, 32)   864         input_3[0][0]                    
__________________________________________________________________________________________________
batch_normalization_94 (BatchNo (None, 74, 74, 32)   96          conv2d_98[0][0]                  
__________________________________________________________________________________________________
activation_94 (Activation)      (None, 74, 74, 32)   0           batch_normalization_94[0][0]     
__________________________________________________________________________________________________
conv2d_99 (Conv2D)              (None, 72, 72, 32)   9216        activation_94[0][0]              
__________________________________________________________________________________________________
batch_normalization_95 (BatchNo (None, 72, 72, 32)   96          conv2d_99[0][0]                  
__________________________________________________________________________________________________
activation_95 (Activation)      (None, 72, 72, 32)   0           batch_normalization_95[0][0]     
__________________________________________________________________________________________________
conv2d_100 (Conv2D)             (None, 72, 72, 64)   18432       activation_95[0][0]              
__________________________________________________________________________________________________
batch_normalization_96 (BatchNo (None, 72, 72, 64)   192         conv2d_100[0][0]                 
__________________________________________________________________________________________________
activation_96 (Activation)      (None, 72, 72, 64)   0           batch_normalization_96[0][0]     
__________________________________________________________________________________________________
max_pooling2d_8 (MaxPooling2D)  (None, 35, 35, 64)   0           activation_96[0][0]              
__________________________________________________________________________________________________
conv2d_101 (Conv2D)             (None, 35, 35, 80)   5120        max_pooling2d_8[0][0]            
__________________________________________________________________________________________________
batch_normalization_97 (BatchNo (None, 35, 35, 80)   240         conv2d_101[0][0]                 
__________________________________________________________________________________________________
activation_97 (Activation)      (None, 35, 35, 80)   0           batch_normalization_97[0][0]     
__________________________________________________________________________________________________
conv2d_102 (Conv2D)             (None, 33, 33, 192)  138240      activation_97[0][0]              
__________________________________________________________________________________________________
batch_normalization_98 (BatchNo (None, 33, 33, 192)  576         conv2d_102[0][0]                 
__________________________________________________________________________________________________
activation_98 (Activation)      (None, 33, 33, 192)  0           batch_normalization_98[0][0]     
__________________________________________________________________________________________________
max_pooling2d_9 (MaxPooling2D)  (None, 16, 16, 192)  0           activation_98[0][0]              
__________________________________________________________________________________________________
conv2d_106 (Conv2D)             (None, 16, 16, 64)   12288       max_pooling2d_9[0][0]            
__________________________________________________________________________________________________
batch_normalization_102 (BatchN (None, 16, 16, 64)   192         conv2d_106[0][0]                 
__________________________________________________________________________________________________
activation_102 (Activation)     (None, 16, 16, 64)   0           batch_normalization_102[0][0]    
__________________________________________________________________________________________________
conv2d_104 (Conv2D)             (None, 16, 16, 48)   9216        max_pooling2d_9[0][0]            
__________________________________________________________________________________________________
conv2d_107 (Conv2D)             (None, 16, 16, 96)   55296       activation_102[0][0]             
__________________________________________________________________________________________________
batch_normalization_100 (BatchN (None, 16, 16, 48)   144         conv2d_104[0][0]                 
__________________________________________________________________________________________________
batch_normalization_103 (BatchN (None, 16, 16, 96)   288         conv2d_107[0][0]                 
__________________________________________________________________________________________________
activation_100 (Activation)     (None, 16, 16, 48)   0           batch_normalization_100[0][0]    
__________________________________________________________________________________________________
activation_103 (Activation)     (None, 16, 16, 96)   0           batch_normalization_103[0][0]    
__________________________________________________________________________________________________
average_pooling2d_9 (AveragePoo (None, 16, 16, 192)  0           max_pooling2d_9[0][0]            
__________________________________________________________________________________________________
conv2d_103 (Conv2D)             (None, 16, 16, 64)   12288       max_pooling2d_9[0][0]            
__________________________________________________________________________________________________
conv2d_105 (Conv2D)             (None, 16, 16, 64)   76800       activation_100[0][0]             
__________________________________________________________________________________________________
conv2d_108 (Conv2D)             (None, 16, 16, 96)   82944       activation_103[0][0]             
__________________________________________________________________________________________________
conv2d_109 (Conv2D)             (None, 16, 16, 32)   6144        average_pooling2d_9[0][0]        
__________________________________________________________________________________________________
batch_normalization_99 (BatchNo (None, 16, 16, 64)   192         conv2d_103[0][0]                 
__________________________________________________________________________________________________
batch_normalization_101 (BatchN (None, 16, 16, 64)   192         conv2d_105[0][0]                 
__________________________________________________________________________________________________
batch_normalization_104 (BatchN (None, 16, 16, 96)   288         conv2d_108[0][0]                 
__________________________________________________________________________________________________
batch_normalization_105 (BatchN (None, 16, 16, 32)   96          conv2d_109[0][0]                 
__________________________________________________________________________________________________
activation_99 (Activation)      (None, 16, 16, 64)   0           batch_normalization_99[0][0]     
__________________________________________________________________________________________________
activation_101 (Activation)     (None, 16, 16, 64)   0           batch_normalization_101[0][0]    
__________________________________________________________________________________________________
activation_104 (Activation)     (None, 16, 16, 96)   0           batch_normalization_104[0][0]    
__________________________________________________________________________________________________
activation_105 (Activation)     (None, 16, 16, 32)   0           batch_normalization_105[0][0]    
__________________________________________________________________________________________________
mixed0 (Concatenate)            (None, 16, 16, 256)  0           activation_99[0][0]              
                                                                 activation_101[0][0]             
                                                                 activation_104[0][0]             
                                                                 activation_105[0][0]             
__________________________________________________________________________________________________
conv2d_113 (Conv2D)             (None, 16, 16, 64)   16384       mixed0[0][0]                     
__________________________________________________________________________________________________
batch_normalization_109 (BatchN (None, 16, 16, 64)   192         conv2d_113[0][0]                 
__________________________________________________________________________________________________
activation_109 (Activation)     (None, 16, 16, 64)   0           batch_normalization_109[0][0]    
__________________________________________________________________________________________________
conv2d_111 (Conv2D)             (None, 16, 16, 48)   12288       mixed0[0][0]                     
__________________________________________________________________________________________________
conv2d_114 (Conv2D)             (None, 16, 16, 96)   55296       activation_109[0][0]             
__________________________________________________________________________________________________
batch_normalization_107 (BatchN (None, 16, 16, 48)   144         conv2d_111[0][0]                 
__________________________________________________________________________________________________
batch_normalization_110 (BatchN (None, 16, 16, 96)   288         conv2d_114[0][0]                 
__________________________________________________________________________________________________
activation_107 (Activation)     (None, 16, 16, 48)   0           batch_normalization_107[0][0]    
__________________________________________________________________________________________________
activation_110 (Activation)     (None, 16, 16, 96)   0           batch_normalization_110[0][0]    
__________________________________________________________________________________________________
average_pooling2d_10 (AveragePo (None, 16, 16, 256)  0           mixed0[0][0]                     
__________________________________________________________________________________________________
conv2d_110 (Conv2D)             (None, 16, 16, 64)   16384       mixed0[0][0]                     
__________________________________________________________________________________________________
conv2d_112 (Conv2D)             (None, 16, 16, 64)   76800       activation_107[0][0]             
__________________________________________________________________________________________________
conv2d_115 (Conv2D)             (None, 16, 16, 96)   82944       activation_110[0][0]             
__________________________________________________________________________________________________
conv2d_116 (Conv2D)             (None, 16, 16, 64)   16384       average_pooling2d_10[0][0]       
__________________________________________________________________________________________________
batch_normalization_106 (BatchN (None, 16, 16, 64)   192         conv2d_110[0][0]                 
__________________________________________________________________________________________________
batch_normalization_108 (BatchN (None, 16, 16, 64)   192         conv2d_112[0][0]                 
__________________________________________________________________________________________________
batch_normalization_111 (BatchN (None, 16, 16, 96)   288         conv2d_115[0][0]                 
__________________________________________________________________________________________________
batch_normalization_112 (BatchN (None, 16, 16, 64)   192         conv2d_116[0][0]                 
__________________________________________________________________________________________________
activation_106 (Activation)     (None, 16, 16, 64)   0           batch_normalization_106[0][0]    
__________________________________________________________________________________________________
activation_108 (Activation)     (None, 16, 16, 64)   0           batch_normalization_108[0][0]    
__________________________________________________________________________________________________
activation_111 (Activation)     (None, 16, 16, 96)   0           batch_normalization_111[0][0]    
__________________________________________________________________________________________________
activation_112 (Activation)     (None, 16, 16, 64)   0           batch_normalization_112[0][0]    
__________________________________________________________________________________________________
mixed1 (Concatenate)            (None, 16, 16, 288)  0           activation_106[0][0]             
                                                                 activation_108[0][0]             
                                                                 activation_111[0][0]             
                                                                 activation_112[0][0]             
__________________________________________________________________________________________________
conv2d_120 (Conv2D)             (None, 16, 16, 64)   18432       mixed1[0][0]                     
__________________________________________________________________________________________________
batch_normalization_116 (BatchN (None, 16, 16, 64)   192         conv2d_120[0][0]                 
__________________________________________________________________________________________________
activation_116 (Activation)     (None, 16, 16, 64)   0           batch_normalization_116[0][0]    
__________________________________________________________________________________________________
conv2d_118 (Conv2D)             (None, 16, 16, 48)   13824       mixed1[0][0]                     
__________________________________________________________________________________________________
conv2d_121 (Conv2D)             (None, 16, 16, 96)   55296       activation_116[0][0]             
__________________________________________________________________________________________________
batch_normalization_114 (BatchN (None, 16, 16, 48)   144         conv2d_118[0][0]                 
__________________________________________________________________________________________________
batch_normalization_117 (BatchN (None, 16, 16, 96)   288         conv2d_121[0][0]                 
__________________________________________________________________________________________________
activation_114 (Activation)     (None, 16, 16, 48)   0           batch_normalization_114[0][0]    
__________________________________________________________________________________________________
activation_117 (Activation)     (None, 16, 16, 96)   0           batch_normalization_117[0][0]    
__________________________________________________________________________________________________
average_pooling2d_11 (AveragePo (None, 16, 16, 288)  0           mixed1[0][0]                     
__________________________________________________________________________________________________
conv2d_117 (Conv2D)             (None, 16, 16, 64)   18432       mixed1[0][0]                     
__________________________________________________________________________________________________
conv2d_119 (Conv2D)             (None, 16, 16, 64)   76800       activation_114[0][0]             
__________________________________________________________________________________________________
conv2d_122 (Conv2D)             (None, 16, 16, 96)   82944       activation_117[0][0]             
__________________________________________________________________________________________________
conv2d_123 (Conv2D)             (None, 16, 16, 64)   18432       average_pooling2d_11[0][0]       
__________________________________________________________________________________________________
batch_normalization_113 (BatchN (None, 16, 16, 64)   192         conv2d_117[0][0]                 
__________________________________________________________________________________________________
batch_normalization_115 (BatchN (None, 16, 16, 64)   192         conv2d_119[0][0]                 
__________________________________________________________________________________________________
batch_normalization_118 (BatchN (None, 16, 16, 96)   288         conv2d_122[0][0]                 
__________________________________________________________________________________________________
batch_normalization_119 (BatchN (None, 16, 16, 64)   192         conv2d_123[0][0]                 
__________________________________________________________________________________________________
activation_113 (Activation)     (None, 16, 16, 64)   0           batch_normalization_113[0][0]    
__________________________________________________________________________________________________
activation_115 (Activation)     (None, 16, 16, 64)   0           batch_normalization_115[0][0]    
__________________________________________________________________________________________________
activation_118 (Activation)     (None, 16, 16, 96)   0           batch_normalization_118[0][0]    
__________________________________________________________________________________________________
activation_119 (Activation)     (None, 16, 16, 64)   0           batch_normalization_119[0][0]    
__________________________________________________________________________________________________
mixed2 (Concatenate)            (None, 16, 16, 288)  0           activation_113[0][0]             
                                                                 activation_115[0][0]             
                                                                 activation_118[0][0]             
                                                                 activation_119[0][0]             
__________________________________________________________________________________________________
conv2d_125 (Conv2D)             (None, 16, 16, 64)   18432       mixed2[0][0]                     
__________________________________________________________________________________________________
batch_normalization_121 (BatchN (None, 16, 16, 64)   192         conv2d_125[0][0]                 
__________________________________________________________________________________________________
activation_121 (Activation)     (None, 16, 16, 64)   0           batch_normalization_121[0][0]    
__________________________________________________________________________________________________
conv2d_126 (Conv2D)             (None, 16, 16, 96)   55296       activation_121[0][0]             
__________________________________________________________________________________________________
batch_normalization_122 (BatchN (None, 16, 16, 96)   288         conv2d_126[0][0]                 
__________________________________________________________________________________________________
activation_122 (Activation)     (None, 16, 16, 96)   0           batch_normalization_122[0][0]    
__________________________________________________________________________________________________
conv2d_124 (Conv2D)             (None, 7, 7, 384)    995328      mixed2[0][0]                     
__________________________________________________________________________________________________
conv2d_127 (Conv2D)             (None, 7, 7, 96)     82944       activation_122[0][0]             
__________________________________________________________________________________________________
batch_normalization_120 (BatchN (None, 7, 7, 384)    1152        conv2d_124[0][0]                 
__________________________________________________________________________________________________
batch_normalization_123 (BatchN (None, 7, 7, 96)     288         conv2d_127[0][0]                 
__________________________________________________________________________________________________
activation_120 (Activation)     (None, 7, 7, 384)    0           batch_normalization_120[0][0]    
__________________________________________________________________________________________________
activation_123 (Activation)     (None, 7, 7, 96)     0           batch_normalization_123[0][0]    
__________________________________________________________________________________________________
max_pooling2d_10 (MaxPooling2D) (None, 7, 7, 288)    0           mixed2[0][0]                     
__________________________________________________________________________________________________
mixed3 (Concatenate)            (None, 7, 7, 768)    0           activation_120[0][0]             
                                                                 activation_123[0][0]             
                                                                 max_pooling2d_10[0][0]           
__________________________________________________________________________________________________
conv2d_132 (Conv2D)             (None, 7, 7, 128)    98304       mixed3[0][0]                     
__________________________________________________________________________________________________
batch_normalization_128 (BatchN (None, 7, 7, 128)    384         conv2d_132[0][0]                 
__________________________________________________________________________________________________
activation_128 (Activation)     (None, 7, 7, 128)    0           batch_normalization_128[0][0]    
__________________________________________________________________________________________________
conv2d_133 (Conv2D)             (None, 7, 7, 128)    114688      activation_128[0][0]             
__________________________________________________________________________________________________
batch_normalization_129 (BatchN (None, 7, 7, 128)    384         conv2d_133[0][0]                 
__________________________________________________________________________________________________
activation_129 (Activation)     (None, 7, 7, 128)    0           batch_normalization_129[0][0]    
__________________________________________________________________________________________________
conv2d_129 (Conv2D)             (None, 7, 7, 128)    98304       mixed3[0][0]                     
__________________________________________________________________________________________________
conv2d_134 (Conv2D)             (None, 7, 7, 128)    114688      activation_129[0][0]             
__________________________________________________________________________________________________
batch_normalization_125 (BatchN (None, 7, 7, 128)    384         conv2d_129[0][0]                 
__________________________________________________________________________________________________
batch_normalization_130 (BatchN (None, 7, 7, 128)    384         conv2d_134[0][0]                 
__________________________________________________________________________________________________
activation_125 (Activation)     (None, 7, 7, 128)    0           batch_normalization_125[0][0]    
__________________________________________________________________________________________________
activation_130 (Activation)     (None, 7, 7, 128)    0           batch_normalization_130[0][0]    
__________________________________________________________________________________________________
conv2d_130 (Conv2D)             (None, 7, 7, 128)    114688      activation_125[0][0]             
__________________________________________________________________________________________________
conv2d_135 (Conv2D)             (None, 7, 7, 128)    114688      activation_130[0][0]             
__________________________________________________________________________________________________
batch_normalization_126 (BatchN (None, 7, 7, 128)    384         conv2d_130[0][0]                 
__________________________________________________________________________________________________
batch_normalization_131 (BatchN (None, 7, 7, 128)    384         conv2d_135[0][0]                 
__________________________________________________________________________________________________
activation_126 (Activation)     (None, 7, 7, 128)    0           batch_normalization_126[0][0]    
__________________________________________________________________________________________________
activation_131 (Activation)     (None, 7, 7, 128)    0           batch_normalization_131[0][0]    
__________________________________________________________________________________________________
average_pooling2d_12 (AveragePo (None, 7, 7, 768)    0           mixed3[0][0]                     
__________________________________________________________________________________________________
conv2d_128 (Conv2D)             (None, 7, 7, 192)    147456      mixed3[0][0]                     
__________________________________________________________________________________________________
conv2d_131 (Conv2D)             (None, 7, 7, 192)    172032      activation_126[0][0]             
__________________________________________________________________________________________________
conv2d_136 (Conv2D)             (None, 7, 7, 192)    172032      activation_131[0][0]             
__________________________________________________________________________________________________
conv2d_137 (Conv2D)             (None, 7, 7, 192)    147456      average_pooling2d_12[0][0]       
__________________________________________________________________________________________________
batch_normalization_124 (BatchN (None, 7, 7, 192)    576         conv2d_128[0][0]                 
__________________________________________________________________________________________________
batch_normalization_127 (BatchN (None, 7, 7, 192)    576         conv2d_131[0][0]                 
__________________________________________________________________________________________________
batch_normalization_132 (BatchN (None, 7, 7, 192)    576         conv2d_136[0][0]                 
__________________________________________________________________________________________________
batch_normalization_133 (BatchN (None, 7, 7, 192)    576         conv2d_137[0][0]                 
__________________________________________________________________________________________________
activation_124 (Activation)     (None, 7, 7, 192)    0           batch_normalization_124[0][0]    
__________________________________________________________________________________________________
activation_127 (Activation)     (None, 7, 7, 192)    0           batch_normalization_127[0][0]    
__________________________________________________________________________________________________
activation_132 (Activation)     (None, 7, 7, 192)    0           batch_normalization_132[0][0]    
__________________________________________________________________________________________________
activation_133 (Activation)     (None, 7, 7, 192)    0           batch_normalization_133[0][0]    
__________________________________________________________________________________________________
mixed4 (Concatenate)            (None, 7, 7, 768)    0           activation_124[0][0]             
                                                                 activation_127[0][0]             
                                                                 activation_132[0][0]             
                                                                 activation_133[0][0]             
__________________________________________________________________________________________________
conv2d_142 (Conv2D)             (None, 7, 7, 160)    122880      mixed4[0][0]                     
__________________________________________________________________________________________________
batch_normalization_138 (BatchN (None, 7, 7, 160)    480         conv2d_142[0][0]                 
__________________________________________________________________________________________________
activation_138 (Activation)     (None, 7, 7, 160)    0           batch_normalization_138[0][0]    
__________________________________________________________________________________________________
conv2d_143 (Conv2D)             (None, 7, 7, 160)    179200      activation_138[0][0]             
__________________________________________________________________________________________________
batch_normalization_139 (BatchN (None, 7, 7, 160)    480         conv2d_143[0][0]                 
__________________________________________________________________________________________________
activation_139 (Activation)     (None, 7, 7, 160)    0           batch_normalization_139[0][0]    
__________________________________________________________________________________________________
conv2d_139 (Conv2D)             (None, 7, 7, 160)    122880      mixed4[0][0]                     
__________________________________________________________________________________________________
conv2d_144 (Conv2D)             (None, 7, 7, 160)    179200      activation_139[0][0]             
__________________________________________________________________________________________________
batch_normalization_135 (BatchN (None, 7, 7, 160)    480         conv2d_139[0][0]                 
__________________________________________________________________________________________________
batch_normalization_140 (BatchN (None, 7, 7, 160)    480         conv2d_144[0][0]                 
__________________________________________________________________________________________________
activation_135 (Activation)     (None, 7, 7, 160)    0           batch_normalization_135[0][0]    
__________________________________________________________________________________________________
activation_140 (Activation)     (None, 7, 7, 160)    0           batch_normalization_140[0][0]    
__________________________________________________________________________________________________
conv2d_140 (Conv2D)             (None, 7, 7, 160)    179200      activation_135[0][0]             
__________________________________________________________________________________________________
conv2d_145 (Conv2D)             (None, 7, 7, 160)    179200      activation_140[0][0]             
__________________________________________________________________________________________________
batch_normalization_136 (BatchN (None, 7, 7, 160)    480         conv2d_140[0][0]                 
__________________________________________________________________________________________________
batch_normalization_141 (BatchN (None, 7, 7, 160)    480         conv2d_145[0][0]                 
__________________________________________________________________________________________________
activation_136 (Activation)     (None, 7, 7, 160)    0           batch_normalization_136[0][0]    
__________________________________________________________________________________________________
activation_141 (Activation)     (None, 7, 7, 160)    0           batch_normalization_141[0][0]    
__________________________________________________________________________________________________
average_pooling2d_13 (AveragePo (None, 7, 7, 768)    0           mixed4[0][0]                     
__________________________________________________________________________________________________
conv2d_138 (Conv2D)             (None, 7, 7, 192)    147456      mixed4[0][0]                     
__________________________________________________________________________________________________
conv2d_141 (Conv2D)             (None, 7, 7, 192)    215040      activation_136[0][0]             
__________________________________________________________________________________________________
conv2d_146 (Conv2D)             (None, 7, 7, 192)    215040      activation_141[0][0]             
__________________________________________________________________________________________________
conv2d_147 (Conv2D)             (None, 7, 7, 192)    147456      average_pooling2d_13[0][0]       
__________________________________________________________________________________________________
batch_normalization_134 (BatchN (None, 7, 7, 192)    576         conv2d_138[0][0]                 
__________________________________________________________________________________________________
batch_normalization_137 (BatchN (None, 7, 7, 192)    576         conv2d_141[0][0]                 
__________________________________________________________________________________________________
batch_normalization_142 (BatchN (None, 7, 7, 192)    576         conv2d_146[0][0]                 
__________________________________________________________________________________________________
batch_normalization_143 (BatchN (None, 7, 7, 192)    576         conv2d_147[0][0]                 
__________________________________________________________________________________________________
activation_134 (Activation)     (None, 7, 7, 192)    0           batch_normalization_134[0][0]    
__________________________________________________________________________________________________
activation_137 (Activation)     (None, 7, 7, 192)    0           batch_normalization_137[0][0]    
__________________________________________________________________________________________________
activation_142 (Activation)     (None, 7, 7, 192)    0           batch_normalization_142[0][0]    
__________________________________________________________________________________________________
activation_143 (Activation)     (None, 7, 7, 192)    0           batch_normalization_143[0][0]    
__________________________________________________________________________________________________
mixed5 (Concatenate)            (None, 7, 7, 768)    0           activation_134[0][0]             
                                                                 activation_137[0][0]             
                                                                 activation_142[0][0]             
                                                                 activation_143[0][0]             
__________________________________________________________________________________________________
conv2d_152 (Conv2D)             (None, 7, 7, 160)    122880      mixed5[0][0]                     
__________________________________________________________________________________________________
batch_normalization_148 (BatchN (None, 7, 7, 160)    480         conv2d_152[0][0]                 
__________________________________________________________________________________________________
activation_148 (Activation)     (None, 7, 7, 160)    0           batch_normalization_148[0][0]    
__________________________________________________________________________________________________
conv2d_153 (Conv2D)             (None, 7, 7, 160)    179200      activation_148[0][0]             
__________________________________________________________________________________________________
batch_normalization_149 (BatchN (None, 7, 7, 160)    480         conv2d_153[0][0]                 
__________________________________________________________________________________________________
activation_149 (Activation)     (None, 7, 7, 160)    0           batch_normalization_149[0][0]    
__________________________________________________________________________________________________
conv2d_149 (Conv2D)             (None, 7, 7, 160)    122880      mixed5[0][0]                     
__________________________________________________________________________________________________
conv2d_154 (Conv2D)             (None, 7, 7, 160)    179200      activation_149[0][0]             
__________________________________________________________________________________________________
batch_normalization_145 (BatchN (None, 7, 7, 160)    480         conv2d_149[0][0]                 
__________________________________________________________________________________________________
batch_normalization_150 (BatchN (None, 7, 7, 160)    480         conv2d_154[0][0]                 
__________________________________________________________________________________________________
activation_145 (Activation)     (None, 7, 7, 160)    0           batch_normalization_145[0][0]    
__________________________________________________________________________________________________
activation_150 (Activation)     (None, 7, 7, 160)    0           batch_normalization_150[0][0]    
__________________________________________________________________________________________________
conv2d_150 (Conv2D)             (None, 7, 7, 160)    179200      activation_145[0][0]             
__________________________________________________________________________________________________
conv2d_155 (Conv2D)             (None, 7, 7, 160)    179200      activation_150[0][0]             
__________________________________________________________________________________________________
batch_normalization_146 (BatchN (None, 7, 7, 160)    480         conv2d_150[0][0]                 
__________________________________________________________________________________________________
batch_normalization_151 (BatchN (None, 7, 7, 160)    480         conv2d_155[0][0]                 
__________________________________________________________________________________________________
activation_146 (Activation)     (None, 7, 7, 160)    0           batch_normalization_146[0][0]    
__________________________________________________________________________________________________
activation_151 (Activation)     (None, 7, 7, 160)    0           batch_normalization_151[0][0]    
__________________________________________________________________________________________________
average_pooling2d_14 (AveragePo (None, 7, 7, 768)    0           mixed5[0][0]                     
__________________________________________________________________________________________________
conv2d_148 (Conv2D)             (None, 7, 7, 192)    147456      mixed5[0][0]                     
__________________________________________________________________________________________________
conv2d_151 (Conv2D)             (None, 7, 7, 192)    215040      activation_146[0][0]             
__________________________________________________________________________________________________
conv2d_156 (Conv2D)             (None, 7, 7, 192)    215040      activation_151[0][0]             
__________________________________________________________________________________________________
conv2d_157 (Conv2D)             (None, 7, 7, 192)    147456      average_pooling2d_14[0][0]       
__________________________________________________________________________________________________
batch_normalization_144 (BatchN (None, 7, 7, 192)    576         conv2d_148[0][0]                 
__________________________________________________________________________________________________
batch_normalization_147 (BatchN (None, 7, 7, 192)    576         conv2d_151[0][0]                 
__________________________________________________________________________________________________
batch_normalization_152 (BatchN (None, 7, 7, 192)    576         conv2d_156[0][0]                 
__________________________________________________________________________________________________
batch_normalization_153 (BatchN (None, 7, 7, 192)    576         conv2d_157[0][0]                 
__________________________________________________________________________________________________
activation_144 (Activation)     (None, 7, 7, 192)    0           batch_normalization_144[0][0]    
__________________________________________________________________________________________________
activation_147 (Activation)     (None, 7, 7, 192)    0           batch_normalization_147[0][0]    
__________________________________________________________________________________________________
activation_152 (Activation)     (None, 7, 7, 192)    0           batch_normalization_152[0][0]    
__________________________________________________________________________________________________
activation_153 (Activation)     (None, 7, 7, 192)    0           batch_normalization_153[0][0]    
__________________________________________________________________________________________________
mixed6 (Concatenate)            (None, 7, 7, 768)    0           activation_144[0][0]             
                                                                 activation_147[0][0]             
                                                                 activation_152[0][0]             
                                                                 activation_153[0][0]             
__________________________________________________________________________________________________
conv2d_162 (Conv2D)             (None, 7, 7, 192)    147456      mixed6[0][0]                     
__________________________________________________________________________________________________
batch_normalization_158 (BatchN (None, 7, 7, 192)    576         conv2d_162[0][0]                 
__________________________________________________________________________________________________
activation_158 (Activation)     (None, 7, 7, 192)    0           batch_normalization_158[0][0]    
__________________________________________________________________________________________________
conv2d_163 (Conv2D)             (None, 7, 7, 192)    258048      activation_158[0][0]             
__________________________________________________________________________________________________
batch_normalization_159 (BatchN (None, 7, 7, 192)    576         conv2d_163[0][0]                 
__________________________________________________________________________________________________
activation_159 (Activation)     (None, 7, 7, 192)    0           batch_normalization_159[0][0]    
__________________________________________________________________________________________________
conv2d_159 (Conv2D)             (None, 7, 7, 192)    147456      mixed6[0][0]                     
__________________________________________________________________________________________________
conv2d_164 (Conv2D)             (None, 7, 7, 192)    258048      activation_159[0][0]             
__________________________________________________________________________________________________
batch_normalization_155 (BatchN (None, 7, 7, 192)    576         conv2d_159[0][0]                 
__________________________________________________________________________________________________
batch_normalization_160 (BatchN (None, 7, 7, 192)    576         conv2d_164[0][0]                 
__________________________________________________________________________________________________
activation_155 (Activation)     (None, 7, 7, 192)    0           batch_normalization_155[0][0]    
__________________________________________________________________________________________________
activation_160 (Activation)     (None, 7, 7, 192)    0           batch_normalization_160[0][0]    
__________________________________________________________________________________________________
conv2d_160 (Conv2D)             (None, 7, 7, 192)    258048      activation_155[0][0]             
__________________________________________________________________________________________________
conv2d_165 (Conv2D)             (None, 7, 7, 192)    258048      activation_160[0][0]             
__________________________________________________________________________________________________
batch_normalization_156 (BatchN (None, 7, 7, 192)    576         conv2d_160[0][0]                 
__________________________________________________________________________________________________
batch_normalization_161 (BatchN (None, 7, 7, 192)    576         conv2d_165[0][0]                 
__________________________________________________________________________________________________
activation_156 (Activation)     (None, 7, 7, 192)    0           batch_normalization_156[0][0]    
__________________________________________________________________________________________________
activation_161 (Activation)     (None, 7, 7, 192)    0           batch_normalization_161[0][0]    
__________________________________________________________________________________________________
average_pooling2d_15 (AveragePo (None, 7, 7, 768)    0           mixed6[0][0]                     
__________________________________________________________________________________________________
conv2d_158 (Conv2D)             (None, 7, 7, 192)    147456      mixed6[0][0]                     
__________________________________________________________________________________________________
conv2d_161 (Conv2D)             (None, 7, 7, 192)    258048      activation_156[0][0]             
__________________________________________________________________________________________________
conv2d_166 (Conv2D)             (None, 7, 7, 192)    258048      activation_161[0][0]             
__________________________________________________________________________________________________
conv2d_167 (Conv2D)             (None, 7, 7, 192)    147456      average_pooling2d_15[0][0]       
__________________________________________________________________________________________________
batch_normalization_154 (BatchN (None, 7, 7, 192)    576         conv2d_158[0][0]                 
__________________________________________________________________________________________________
batch_normalization_157 (BatchN (None, 7, 7, 192)    576         conv2d_161[0][0]                 
__________________________________________________________________________________________________
batch_normalization_162 (BatchN (None, 7, 7, 192)    576         conv2d_166[0][0]                 
__________________________________________________________________________________________________
batch_normalization_163 (BatchN (None, 7, 7, 192)    576         conv2d_167[0][0]                 
__________________________________________________________________________________________________
activation_154 (Activation)     (None, 7, 7, 192)    0           batch_normalization_154[0][0]    
__________________________________________________________________________________________________
activation_157 (Activation)     (None, 7, 7, 192)    0           batch_normalization_157[0][0]    
__________________________________________________________________________________________________
activation_162 (Activation)     (None, 7, 7, 192)    0           batch_normalization_162[0][0]    
__________________________________________________________________________________________________
activation_163 (Activation)     (None, 7, 7, 192)    0           batch_normalization_163[0][0]    
__________________________________________________________________________________________________
mixed7 (Concatenate)            (None, 7, 7, 768)    0           activation_154[0][0]             
                                                                 activation_157[0][0]             
                                                                 activation_162[0][0]             
                                                                 activation_163[0][0]             
__________________________________________________________________________________________________
conv2d_170 (Conv2D)             (None, 7, 7, 192)    147456      mixed7[0][0]                     
__________________________________________________________________________________________________
batch_normalization_166 (BatchN (None, 7, 7, 192)    576         conv2d_170[0][0]                 
__________________________________________________________________________________________________
activation_166 (Activation)     (None, 7, 7, 192)    0           batch_normalization_166[0][0]    
__________________________________________________________________________________________________
conv2d_171 (Conv2D)             (None, 7, 7, 192)    258048      activation_166[0][0]             
__________________________________________________________________________________________________
batch_normalization_167 (BatchN (None, 7, 7, 192)    576         conv2d_171[0][0]                 
__________________________________________________________________________________________________
activation_167 (Activation)     (None, 7, 7, 192)    0           batch_normalization_167[0][0]    
__________________________________________________________________________________________________
conv2d_168 (Conv2D)             (None, 7, 7, 192)    147456      mixed7[0][0]                     
__________________________________________________________________________________________________
conv2d_172 (Conv2D)             (None, 7, 7, 192)    258048      activation_167[0][0]             
__________________________________________________________________________________________________
batch_normalization_164 (BatchN (None, 7, 7, 192)    576         conv2d_168[0][0]                 
__________________________________________________________________________________________________
batch_normalization_168 (BatchN (None, 7, 7, 192)    576         conv2d_172[0][0]                 
__________________________________________________________________________________________________
activation_164 (Activation)     (None, 7, 7, 192)    0           batch_normalization_164[0][0]    
__________________________________________________________________________________________________
activation_168 (Activation)     (None, 7, 7, 192)    0           batch_normalization_168[0][0]    
__________________________________________________________________________________________________
conv2d_169 (Conv2D)             (None, 3, 3, 320)    552960      activation_164[0][0]             
__________________________________________________________________________________________________
conv2d_173 (Conv2D)             (None, 3, 3, 192)    331776      activation_168[0][0]             
__________________________________________________________________________________________________
batch_normalization_165 (BatchN (None, 3, 3, 320)    960         conv2d_169[0][0]                 
__________________________________________________________________________________________________
batch_normalization_169 (BatchN (None, 3, 3, 192)    576         conv2d_173[0][0]                 
__________________________________________________________________________________________________
activation_165 (Activation)     (None, 3, 3, 320)    0           batch_normalization_165[0][0]    
__________________________________________________________________________________________________
activation_169 (Activation)     (None, 3, 3, 192)    0           batch_normalization_169[0][0]    
__________________________________________________________________________________________________
max_pooling2d_11 (MaxPooling2D) (None, 3, 3, 768)    0           mixed7[0][0]                     
__________________________________________________________________________________________________
mixed8 (Concatenate)            (None, 3, 3, 1280)   0           activation_165[0][0]             
                                                                 activation_169[0][0]             
                                                                 max_pooling2d_11[0][0]           
__________________________________________________________________________________________________
flatten_6 (Flatten)             (None, 11520)        0           mixed8[0][0]                     
__________________________________________________________________________________________________
dense_12 (Dense)                (None, 512)          5898752     flatten_6[0][0]                  
__________________________________________________________________________________________________
dropout_5 (Dropout)             (None, 512)          0           dense_12[0][0]                   
__________________________________________________________________________________________________
dense_13 (Dense)                (None, 1)            513         dropout_5[0][0]                  
==================================================================================================
Total params: 16,574,113
Trainable params: 16,552,737
Non-trainable params: 21,376
__________________________________________________________________________________________________

Data Augmentation

from tensorflow.keras.preprocessing.image import ImageDataGenerator

# Create run-time augmentation on training and test dataset
# For training datagenerator, we add normalization and validation_split
train_datagen = ImageDataGenerator(
    rescale = 1 / 255.0,
    rotation_range = 45,
    width_shift_range = 0.3,
    height_shift_range = 0.3,
    shear_range = 0.3,
    zoom_range = 0.3,
    horizontal_flip = True,
    fill_mode = 'nearest',
    validation_split = 0.1112)

# For test datagenerator, we only normalize the data.
test_datagen = ImageDataGenerator(rescale=1 / 255.0)

# Flow training images in batches of 128 using train_datagen generator
training_generator = train_datagen.flow_from_directory(
    TRAINING_DIR,                # This is the source directory for training images
    target_size = (150, 150), # All images will be resized to 300x300
    batch_size = 20,
    subset = 'training',
    class_mode = 'binary')    # Since we use binary_crossentropy loss, we need binary labels

# Flow training images in batches of 128 using train_datagen generator
validation_generator = train_datagen.flow_from_directory(
    TRAINING_DIR,  # This is the source directory for training images
    target_size = (150, 150),           # All images will be resized to 150x150
    batch_size = 20,
    subset = 'validation',
    class_mode ='binary')  # Since we use binary_crossentropy loss, we need binary labels

# Flow training images in batches of 128 using train_datagen generator
testing_generator = test_datagen.flow_from_directory(
    TESTING_DIR,  # This is the source directory for training images
    target_size = (150, 150),           # All images will be resized to 150x150
    batch_size = 20,
    class_mode ='binary')  # Since we use binary_crossentropy loss, we need binary labels

Found 2400 images belonging to 2 classes.
Found 300 images belonging to 2 classes.
Found 300 images belonging to 2 classes.

#using early stopping to exit training if validation loss is not decreasing even after certain epochs (patience)
earlystopping = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience = 3)

#save the best model with lower validation loss
#checkpointer = ModelCheckpoint(filepath="weights.hdf5", verbose=1, save_best_only=True)

# Using Callback to control training
class myCallback(tf.keras.callbacks.Callback):
    def on_epoch_end(self, epoch, log={}):
        if(log.get('val_accuracy') > 0.96):
            print("\nReached {}% val_accuracy so cancelling training!".format(round(100 * log.get('val_accuracy'), 2)))
            self.model.stop_training = True
            
# callback instance
callbacks = myCallback()

Training

history3 = model.fit(
    training_generator,
    steps_per_epoch = training_generator.n // 20,
    epochs = 10,
    validation_data = validation_generator,
    validation_steps = validation_generator.n // 20,
    callbacks = [callbacks, earlystopping],
    verbose = 2
)

Epoch 1/10
120/120 - 413s - loss: 0.4503 - accuracy: 0.7925 - val_loss: 0.3338 - val_accuracy: 0.8867
Epoch 2/10
120/120 - 413s - loss: 0.2744 - accuracy: 0.8858 - val_loss: 0.2382 - val_accuracy: 0.9067
Epoch 3/10
120/120 - 415s - loss: 0.2383 - accuracy: 0.9038 - val_loss: 0.7355 - val_accuracy: 0.8967
Epoch 4/10
120/120 - 413s - loss: 0.2259 - accuracy: 0.9146 - val_loss: 0.5365 - val_accuracy: 0.9067
Epoch 5/10
120/120 - 412s - loss: 0.1887 - accuracy: 0.9296 - val_loss: 0.3835 - val_accuracy: 0.9367
Epoch 00005: early stopping

#load weights
#model.load_weights("weights_.hdf5")

graph_plot(history3)

Task #9 : Assess the performance of the trained model

• Evaluate the performance of the model on data test
• Evaluate score accuracy on data test
• Visualizing the results
• Print out the classification report and plot matrix confusion

# Evaluate the performance of the model
evaluate = model.evaluate(testing_generator, steps = testing_generator.n // 20, verbose = 2)

print('Accuracy Test : {}'.format(evaluate[1]))

15/15 - 8s - loss: 0.1496 - accuracy: 0.9700
Accuracy Test : 0.9700000286102295

# Check the number of images in the dataset
test = []
label = []

# os.listdir returns the list of files in the folder, in this case image class names
for i in  os.listdir('./testing'):
  test_class = os.listdir(os.path.join('testing', i))
  for j in test_class:
    img = os.path.join('./testing', i, j)
    test.append(img)
    label.append(i)

print('Number of train images : {} \n'.format(len(test)))

Number of train images : 300 

# Create Dataframe
test_df = pd.DataFrame({'Image': test, 'Labels': label})
test_df

	Image	Labels
0	./testing\cats\cat.101.jpg	cats
1	./testing\cats\cat.112.jpg	cats
2	./testing\cats\cat.115.jpg	cats
3	./testing\cats\cat.116.jpg	cats
4	./testing\cats\cat.124.jpg	cats
...	...	...
295	./testing\dogs\dog.958.jpg	dogs
296	./testing\dogs\dog.96.jpg	dogs
297	./testing\dogs\dog.963.jpg	dogs
298	./testing\dogs\dog.97.jpg	dogs
299	./testing\dogs\dog.997.jpg	dogs

300 rows × 2 columns

from PIL.Image import open
from sklearn.metrics import confusion_matrix, classification_report, accuracy_score

prediction = []
original = []
image = []
count = 0
label = {0:'cats', 1: 'dogs'}
a = []
b = []

for item in range(len(test_df)):
    # code to open the image
    img= open(test_df['Image'].tolist()[item])
    
    # resizing the image to (150,150)
    img = img.resize((150, 150))
    
    # appending image to the image list
    image.append(img)
    
    # converting image to array
    img = np.asarray(img, dtype= np.float32)
    
    # normalizing the image
    img = img / 255.0
    
    # reshaping the image in to a 4D array
    img = img.reshape(-1, 150, 150, 3)
    
    # making prediction of the model
    predict = model.predict(img)
    a.append(predict[0][0])
    
    # getting the index corresponding to the highest value in the prediction
    #index_predict = np.argmax(predict)    use it if the activate function is softmax
    index_predict = 0 if predict[0][0] < 0.5 else 1
    b.append(index_predict)
    
    # appending the predicted class to the list
    prediction.append(label[index_predict])
    
    # appending original class to the list
    original.append(test_df['Labels'].tolist()[item])
    

results = pd.DataFrame({'Original': original, 'Prediction': prediction, 'Probability': a, 'Class': b})
results

	Original	Prediction	Probability	Class
0	cats	cats	1.968018e-18	0
1	cats	cats	2.564942e-09	0
2	cats	cats	3.059364e-28	0
3	cats	cats	6.273518e-07	0
4	cats	cats	4.219905e-13	0
...	...	...	...	...
295	dogs	dogs	1.000000e+00	1
296	dogs	dogs	1.000000e+00	1
297	dogs	dogs	9.662517e-01	1
298	dogs	dogs	1.000000e+00	1
299	dogs	dogs	1.000000e+00	1

300 rows × 4 columns

# Getting the test accuracy 
score = accuracy_score(original, prediction)
print("Test Accuracy : {}".format(score))

Test Accuracy : 0.9766666666666667

# Visualizing the results
import random
plt.figure(figsize = (50, 50))
for i in range(40):
    j = random.randint(0,len(image))
    plt.subplot(10, 4, i+1)
    col = 'green' if original[j] == prediction[j] else 'red'
    plt.xlabel("Original: " + original[j] + "  Prediction: " + prediction[j], fontsize = 30, color = col)
    plt.xticks([])
    plt.yticks([])
    plt.imshow(image[j])
plt.show()

# Print out the classification report
print(classification_report(np.asarray(original), np.asarray(prediction)))

              precision    recall  f1-score   support

        cats       0.97      0.98      0.98       150
        dogs       0.98      0.97      0.98       150

    accuracy                           0.98       300
   macro avg       0.98      0.98      0.98       300
weighted avg       0.98      0.98      0.98       300

import seaborn as sns
# plot the confusion matrix
plt.figure(figsize = (5,5))
cm = confusion_matrix(np.asarray(original), np.asarray(prediction))
ax = plt.subplot()
sns.heatmap(cm, annot = True, ax = ax)

ax.set_xlabel('Predicted')
ax.set_ylabel('Original')
ax.set_title('Confusion_matrix')

Text(0.5, 1.0, 'Confusion_matrix')